class: center, middle, inverse, title-slide #
##
https://rpubs.com/jaortega/vieRnesEncuestas
###
--- background-image: url(assets/img/viernesdemarzo.png) background-size: 80% ??? Image credit: [http://cursoronline.usal.es/viernes.html) --- class: inverse, center, middle # Introducción --- # ¿Qué hace especial el trabajar con encuestas? - Las encuestas incluyen una serie de *items* sobre un conjunto de *observaciones*: Parecen corresponderse con los [datos ordenados (*tibbles* o *data.frames*)](https://bookdown.org/content/3515/tidydata.html) -- Dos características los diferencian: - En las encuestas son básicos los **metadatos** para poder trabajar con ellas, en particular los **Metadatos de variable**: - *Etiquetas de variable*: Pregunta en el cuestionario, descripción de la variable. - *Etiquetas de valor*: Redacción de las posibles respuestas. Tipos de no respuesta. -- - Las encuestas rara vez son muestras aleatorias (*encuestas complejas*). Si las tratamos como si fuera m.a. los resultados no son representativos: - Las medias y porcentajes no se corresponden con los de la población. - Los errores estándar no están bien calculados. --- # Soluciones - En las encuestas son básicos los **metadatos** para poder trabajar con ellas, en particular los **Metadatos de variable**: - *Etiquetas de variable*: Pregunta en el cuestionario, descripción de la variable. - *Etiquetas de valor*: Redacción de las posibles respuestas. Tipos de no respuesta. #### [labelled data](https://cran.r-project.org/web/packages/sjlabelled/vignettes/intro_sjlabelled.html) -- - Las encuestas rara vez son muestras aleatorias (*encuestas complejas*). Si lo tratamos como si fuera m.a. los resultados no son representativos: - Las medias y porcentajes no se corresponden con los de la población. -- #### [Uso de pesos en tablas y modelos](https://rpubs.com/jaortega/EncuestaR1) -- - Los errores estándar no están bien calculados. #### [Paquete {survey}](https://rpubs.com/jaortega/EncuestaR2) --- ## Resumen de paquetes útiles para trabajar con encuestas: |Paquete | ¿Para qué? | | :------------- |:------------------------------:| | [{haven}](https://cran.r-project.org/web/packages/haven/) | Importar con etiquetas | | [{labelled}](https://cran.r-project.org/web/packages/labelled/) | Manejar etiquetas y variables. Transformar. | | [{sjlabelled}](https://cran.r-project.org/web/packages/sjlabelled/) | Manejar etiquetas. Transformar variables | | [{sjmisc}](https://cran.r-project.org/web/packages/sjmisc/) | Describir variables. Tablas. | | [{expss}](https://cran.r-project.org/web/packages/expss/) | Manejo variables. Presentar resultados. | | [{memisc}](https://cran.r-project.org/web/packages/memisc/) | Importar, transformar, presentar. | | [{sjPlot}](https://cran.r-project.org/web/packages/sjPlot/) | Tablas, gráficos publicables | | [{survey}](https://r-survey.r-forge.r-project.org/survey/) | Análisis encuestas complejas | --- class: inverse, center, middle # Lectura de datos de encuesta --- .pull-left[ ### Formatos más habituales: - `csv` o `fwf` + Diseño de registro en `xls`: En desuso. [`{microdatosES}`](https://cran.r-project.org/web/packages/MicroDatosEs/index.html) - Formatos internos de R: - `rds`: `data=readRDS(...)` - `RData`: `load(`xxx.RData`) - Formatos específicos de software estadístico: - `sav`: *SPSS* - `dat`: *STATA* - *SAS* ] .pull-right[ ### Lectura de datos SPSS |Paquete | Función | Resultado | | :------------- |:-------------:|:-----------------:| | foreign | `read.spss` | *list* o *data.frame* con atributo *labels*| | haven | `read_sav` | *tibble* clase *haven_labelled* | | sjlabelled | `read_spss` | *tibble* clase *labelled* | | memisc | `spss.system.file` + `as.data.set`| `data.set` convertible a `data.frame` | | eatGADS | `import_spss` | *GADSdat* multinivel | ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) <hr> **Bajar datos** <hr> Variables de interés <hr> Lectura <hr> <br> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ Vamos a realizar un análisis completo (factores en el uso de anticonceptivos) de una encuesta compleja. Empezamos por bajar los datos a nuestro disco. ```r library(tidyverse) urlbase="http://iinei.inei.gob.pe/iinei/srienaho/descarga/SPSS/" # Generamos una tiabla con las extensiones y # contenidos de los ficheros ficheros= tribble(~contenido,~fichero, "DatosBasicos","691-Modulo66.zip", "Nacimientos","691-Modulo67.zip", "Nupcialidad","691-Modulo71.zip" ) # Bajamos a la carpeta data aprovechando las # funciones del paquete purrr del tidyverse ficheros %>% pwalk(~download.file(paste0(urlbase,.y), destfile=paste0("data/",.x,".zip"),mode="wb")) ``` Recordatorio: `%>%`, el operador *después*, define operaciones de forma secuencial. En RStudio: `CTR+MAYUSC+M`. ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) <hr> Bajar datos <hr> **Variables de interés** <hr> Lectura <hr> <br> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[Código] ```r varsel=tribble(~Variable,~Significado,~Módulo,~Fichero,~Tipo, "V313","Uso actual de método","RE223132","Nacimientos","Explicada", "V218","Hijos vivos","RE223132","Nacimientos","Explicativa", "V106","Nivel educativo","REC0111","DatosBasicos","Explicativa", "V012","Edad actual","REC0111","DatosBasicos","Explicativa", "V013","Grupo de edad","REC0111","DatosBasicos","Explicativa", "V025","Lugar de residencia","REC0111","DatosBasicos","Explicativa", "V501","Estado civil","RE516171","Nupcialidad","Explicativa", "V213","Embarazada","RE223132","Nacimientos","Selección", "V536","Actividad sexual","RE516171","Nupcialidad","Selección", "CASEID","Identificación indiv","Todos","Todos","Técnica", "V022","Estrato","REC0111","DatosBasicos","Técnica", "V004","Unidad de área final","REC0111","DatosBasicos","Técnica", "V005","Ponderación (mujer)","REC0111","DatosBasicos","Técnica") ``` ] .panel[.panel-name[Tabla] <table> <thead> <tr> <th style="text-align:left;"> Variable </th> <th style="text-align:left;"> Significado </th> <th style="text-align:left;"> Fichero </th> <th style="text-align:left;"> Tipo </th> </tr> </thead> <tbody> <tr> <td style="text-align:left;"> V313 </td> <td style="text-align:left;"> Uso actual de método </td> <td style="text-align:left;"> Nacimientos </td> <td style="text-align:left;"> Explicada </td> </tr> <tr> <td style="text-align:left;"> V218 </td> <td style="text-align:left;"> Hijos vivos </td> <td style="text-align:left;"> Nacimientos </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V106 </td> <td style="text-align:left;"> Nivel educativo </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V012 </td> <td style="text-align:left;"> Edad actual </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V013 </td> <td style="text-align:left;"> Grupo de edad </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V025 </td> <td style="text-align:left;"> Lugar de residencia </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V501 </td> <td style="text-align:left;"> Estado civil </td> <td style="text-align:left;"> Nupcialidad </td> <td style="text-align:left;"> Explicativa </td> </tr> <tr> <td style="text-align:left;"> V213 </td> <td style="text-align:left;"> Embarazada </td> <td style="text-align:left;"> Nacimientos </td> <td style="text-align:left;"> Selección </td> </tr> <tr> <td style="text-align:left;"> V536 </td> <td style="text-align:left;"> Actividad sexual </td> <td style="text-align:left;"> Nupcialidad </td> <td style="text-align:left;"> Selección </td> </tr> <tr> <td style="text-align:left;"> CASEID </td> <td style="text-align:left;"> Identificación indiv </td> <td style="text-align:left;"> Todos </td> <td style="text-align:left;"> Técnica </td> </tr> <tr> <td style="text-align:left;"> V022 </td> <td style="text-align:left;"> Estrato </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Técnica </td> </tr> <tr> <td style="text-align:left;"> V004 </td> <td style="text-align:left;"> Unidad de área final </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Técnica </td> </tr> <tr> <td style="text-align:left;"> V005 </td> <td style="text-align:left;"> Ponderación (mujer) </td> <td style="text-align:left;"> DatosBasicos </td> <td style="text-align:left;"> Técnica </td> </tr> </tbody> </table> ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) <hr> Bajar datos <hr> Variables de interés <hr> **Lectura** <hr> <br> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[Descomprimimos] ```r temp <- tempfile() varsel %>% filter(Fichero!="Todos") %>% pull(Fichero) %>% unique() %>% paste0("data/",.,".zip") %>% walk(unzip,junkpaths=TRUE,exdir=temp) #Descomprime todo en la misma carpeta .m = varsel %>% filter(Módulo!="Todos") %>% pull(Módulo) %>% unique() dir(temp) %>% as_tibble() %>% filter(value %>% str_detect(.m %>% paste(collapse="|"))) %>% mutate(value=paste(temp,value,sep="/")) %>% pull(value) %>% file.copy("data/") rm(.m) ``` Las funciones del [**{tidyverse}**](https://www.tidyverse.org/) son utilísimas para manipular *tiablas* Esta encuesta está bien definida, y {haven} funciona perfectamente. [Aquí](https://rpubs.com/jaortega/EncuestaR1) podéis ver otro ejemplo, con una encuesta del INE que tiene variables con problemas y hay que utilizar otros alternativas de lectura. ] .panel[.panel-name[Lectura con {haven}] Leemos directamente desde el `sav` el archivo de datos básicos, al que le pegamos los datos de nacimientos y nupcialidad. Este es un caso complejo porque los datos están en varios ficheros. Si sólo hay uno es muy sencillo. ```r library(haven) DHS=read_sav("data/REC0111.sav") %>% left_join(read_sav("data/RE223132.sav"),by=c("ID1","CASEID")) %>% left_join(read_sav("data/RE516171.sav"),by=c("ID1","CASEID")) class(DHS) # Se trata de una "tiabla" ``` ``` ## [1] "tbl_df" "tbl" "data.frame" ``` ```r dim(DHS) # Con estas dimensiones: ``` ``` ## [1] 38335 361 ``` ] ] ] --- class: inverse, center, middle # Trabajando con etiquetas: *labelled data* --- ## Objetos `labelled` en R - Hay dos [clases en R con etiquetas](https://cran.r-project.org/web/packages/sjlabelled/vignettes/intro_sjlabelled.html), compatibles entre sí, `labelled` y `haven-labelled`. - Ambas tienen etiquetas de variable (*variable labels*) y etiquetas de valores (*value labels*). - Desafortunadamente la mayoría de los paquetes de R no *entienden* las variables con etiquetas. - Aparentemente son variables numéricas, pero cada valor tiene asignado un código. - Podemos trabajar con los paquetes que entienden las etiquetas, o convertirlas a factores o variables numéricas. ![](data:image/png;base64,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) .footnote[ [Fuente: {labelled}](https://cran.r-project.org/web/packages/labelled/vignettes/intro_labelled.html)] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) <br> <hr> ### Busqueda de variables <hr <br> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[`{sjmisc}`] ```r # install.packages("sjmisc") DHS %>% sjmisc::find_var("Edad") %>% knitr::kable() ``` | col.nr|var.name |var.label | |------:|:--------|:------------------------------------------| | 13|V012 |Edad actual - entrevistada | | 14|V013 |Edad actual por grupos de 5 años | | 75|V152 |Edad del jefe del hogar | | 117|V212 |Edad del entrevistado al primer nacimiento | | 159|V320 |Edad en la esterilización | | 290|V511 |Edad al primer matrimonio | | 293|V525 |Edad en la primera relación sexual | | 298|V531 |Edad en la primera relación sexual (imp) | | 344|V730 |Edad del esposo/compañero | ] .panel[.panel-name[`{labelled}`] ```r # install.packages("labelled") labelled::look_for(DHS,"hijos") %>% knitr::kable() ``` | pos|variable |label |col_type |class |type |levels |value_labels |na_values |na_range | unique_values| n_na|range | |---:|:--------|:-------------------------------------------------------------------------------------------|:--------|:--------------------------------------|:------|:------|:-------------------|:---------|:--------|-------------:|-----:|:-----| | 107|V202 |¿Cuántos hijos viven con usted? |dbl |numeric |double |NULL |NULL |NULL |NULL | 9| 1413|0, 7 | | 109|V204 |¿Cuántos hijos no están viviendo con usted? |dbl |numeric |double |NULL |NULL |NULL |NULL | 8| 1413|0, 6 | | 111|V206 |¿Cuántos hijos han muerto? |dbl |numeric |double |NULL |NULL |NULL |NULL | 6| 1413|0, 4 | | 124|V219 |Número total de hijos vivos incluido el embarazo actual |dbl |numeric |double |NULL |NULL |NULL |NULL | 15| 1413|0, 13 | | 125|V220 |Número total de hijos vivos incluido el embarazo actual (6, si es mayor de 6) |dbl+lbl |haven_labelled, vctrs_vctr , double |double |NULL |6 |NULL |NULL | 8| 1413|0, 6 | | 150|V310 |Cuantas hijas e hijos tenía cuando empezó a usar el primer método para no quedar embarazada |dbl |numeric |double |NULL |NULL |NULL |NULL | 13| 9415|0, 12 | | 151|V311 |Cuantas hijas e hijos tenía cuando empezó a usar el primer método (Grupos) |dbl+lbl |haven_labelled, vctrs_vctr , double |double |NULL |4, 5 |NULL |NULL | 7| 5024|0, 5 | | 310|V605 |Deseo de tener más hijos |dbl+lbl |haven_labelled, vctrs_vctr , double |double |NULL |1, 2, 3, 4, 5, 6, 7 |NULL |NULL | 8| 5024|1, 7 | | 314|V621 |Piensa que su esposo/compañero desea el mismo número de hijos |dbl+lbl |haven_labelled, vctrs_vctr , double |double |NULL |1, 2, 3, 8 |NULL |NULL | 5| 18367|1, 8 | | 319|V627 |Número ideal de hijos |dbl+lbl |haven_labelled, vctrs_vctr , double |double |NULL |96 |NULL |NULL | 14| 5024|0, 96 | ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Describir variables <hr> **Funciones genéricas** <hr> Etiquetas <hr> Tabulaciones <hr> Estadísticos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ Seleccionamos nuestra variable de interés, `V313`, la variable de uso de anticonceptivos. Obviamente no es una variable cuantitativa, pero los datos `labelled` están almacenados como número, con las etiquetas como atributos. ```r class(DHS$V313) #Clase del objeto ``` ``` ## [1] "haven_labelled" "vctrs_vctr" "double" ``` ```r summary(DHS$V313) #Resumen. Función genérica, pero sin método para labelled ``` ``` ## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's ## 0.000 0.000 2.000 1.673 3.000 3.000 5024 ``` > Aunque no sea numérico, `summary` *no se da cuenta*. Eso les pasa a muchas funciones. #### Es básico trabajar con funciones que entiendan los datos `labelled` o transformarlas a factores o variables cuantitativas. ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Describir variables <hr> Funciones genéricas <hr> **Etiquetas** <hr> Tabulaciones <hr> Estadísticos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ Tanto `sjlabelled` como `labelled` y `expss` tienen funciones. .panelset[ .panel[.panel-name[`{sjlabelled}`] ```r # install.packages("sjlabelled") sjlabelled::get_label(DHS$V313) #Etiqueta de variable ``` ``` ## [1] "Uso actual por tipo de método" ``` ```r sjlabelled::get_labels(DHS$V313) #Etiquetas de valor ``` ``` ## [1] "No hay método" "Método folclórico" ## [3] "Método tradicional" "Método moderno" ``` ] .panel[.panel-name[`{labelled}`] ```r # install.packages("labelled") labelled::var_label(DHS$V313) #Etiqueta de variable ``` ``` ## [1] "Uso actual por tipo de método" ``` ```r labelled::val_labels(DHS$V313) # Etiquetas de valor ``` ``` ## No hay método Método folclórico Método tradicional ## 0 1 2 ## Método moderno ## 3 ``` ] .panel[.panel-name[`{expss}`] ```r # install.packages("expss") expss::var_lab(DHS$V313) #Etiqueta de variable ``` ``` ## [1] "Uso actual por tipo de método" ``` ```r expss::val_lab(DHS$V313) # Etiquetas de valor ``` ``` ## No hay método Método folclórico Método tradicional ## 0 1 2 ## Método moderno ## 3 ``` ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Describir variables <hr> Funciones genéricas <hr> Etiquetas <hr> **Tabulaciones** <hr> Estadísticos <hr> .footnote[ Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ Tanto `sjmisc` como `memisc` o el `tidyverse` tienen funciones. .panelset[ .panel[.panel-name[{sjmisc}] ```r # install.packages("sjmisc") sjmisc::frq(DHS$V313) #Tabulación simple ``` ``` ## ## Uso actual por tipo de método (x) <numeric> ## # total N=38335 valid N=33311 mean=1.67 sd=1.39 ## ## Value | Label | N | Raw % | Valid % | Cum. % ## ------------------------------------------------------------- ## 0 | No hay método | 13017 | 33.96 | 39.08 | 39.08 ## 1 | Método folclórico | 188 | 0.49 | 0.56 | 39.64 ## 2 | Método tradicional | 4762 | 12.42 | 14.30 | 53.94 ## 3 | Método moderno | 15344 | 40.03 | 46.06 | 100.00 ## <NA> | <NA> | 5024 | 13.11 | <NA> | <NA> ``` ] .panel[.panel-name[{memisc}] ```r # install.packages("memisc") memisc::codebook(DHS$V313) # Si hay más variables, de todas. ``` ``` ## ============================================================ ## ## DHS$V313 'Uso actual por tipo de método' ## ## ------------------------------------------------------------ ## ## Storage mode: double ## Measurement: undefined ## ## Values and labels N Valid Total ## ## 0 'No hay método' 13017 39.1 34.0 ## 1 'Método folclórico' 188 0.6 0.5 ## 2 'Método tradicional' 4762 14.3 12.4 ## 3 'Método moderno' 15344 46.1 40.0 ## NA M 5024 13.1 ``` ] .panel[.panel-name[{tidyverse}] Las versiones antiguas no reconocían etiquetas. ¡Ahora sí! ```r DHS %>% count(V313) %>% mutate(`% (todos)`=n/sum(n)*100, `% (validos)`=n/(sum(n)-last(n))*100) ``` ``` ## # A tibble: 5 x 4 ## V313 n `% (todos)` `% (validos)` ## * <dbl+lbl> <int> <dbl> <dbl> ## 1 0 [No hay método] 13017 34.0 39.1 ## 2 1 [Método folclórico] 188 0.490 0.564 ## 3 2 [Método tradicional] 4762 12.4 14.3 ## 4 3 [Método moderno] 15344 40.0 46.1 ## 5 NA 5024 13.1 15.1 ``` ] ] ] --- .left-column[ ##Ejemplo: ### Describir variables cuantitativas <hr> Funciones genéricas <hr> Etiquetas <hr> Tabulaciones <hr> **Estadísticos** <hr> <br> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2) ] ] .right-column[ .panelset[ .panel[.panel-name[{sjmisc}] ```r sjmisc::descr(DHS %>% select(V201,V218)) ``` ``` ## ## ## Basic descriptive statistics ## ## var type label n NA.prc mean sd ## V201 numeric Total de niños nacidos 36922 3.69 1.78 1.75 ## V218 numeric Número de niños vivos 36922 3.69 1.73 1.67 ## se md trimmed range iqr skew ## 0.01 1 1.54 13 (0-13) 3 1.33 ## 0.01 1 1.51 13 (0-13) 3 1.23 ``` ] .panel[.panel-name[{Hmisc}] ```r # install.packages("Hmisc") Hmisc::describe(DHS %>% select(V201)) # Si hay más variables, de todas. ``` ``` ## DHS %>% select(V201) ## ## 1 Variables 38335 Observations ## ------------------------------------------------------------ ## V201 : Total de niños nacidos Format:F2.0 ## n missing distinct Info Mean Gmd ## 36922 1413 14 0.953 1.782 1.837 ## .05 .10 .25 .50 .75 .90 ## 0 0 0 1 3 4 ## .95 ## 5 ## ## lowest : 0 1 2 3 4, highest: 9 10 11 12 13 ## ## Value 0 1 2 3 4 5 6 7 ## Frequency 10565 7950 8074 5099 2583 1294 642 328 ## Proportion 0.286 0.215 0.219 0.138 0.070 0.035 0.017 0.009 ## ## Value 8 9 10 11 12 13 ## Frequency 187 99 66 22 10 3 ## Proportion 0.005 0.003 0.002 0.001 0.000 0.000 ## ------------------------------------------------------------ ``` ] .panel[.panel-name[{qwraps2}] ```r # install.packages("qwraps2") options(qwraps2_markup = "markdown") qwraps2::summary_table(DHS %>% select(V201,V218)) ``` | |DHS %>% select(V201, V218) (N = 38,335) | |:----------------------------|:---------------------------------------| |**Total de niños nacidos** | | | minimum |0.00 | | median (IQR) |1.00 (0.00, 3.00) | | mean (sd) |1.78 ± 1.75 | | maximum |13.00 | | Unknown/Missing |1,413 (3.69%) | |**Número de niños vivos** | | | minimum |0.00 | | median (IQR) |1.00 (0.00, 3.00) | | mean (sd) |1.73 ± 1.67 | | maximum |13.00 | | Unknown/Missing |1,413 (3.69%) | ] ] ] --- class: inverse, center, middle # Encuestas complejas: Uso de pesos en tabulaciones y modelos --- ## ¿Por qué es necesario usar pesos? La gran mayoría de encuestas NO son una **muestra aleatoria simple**. Suelen ser muestreo aleatorio dentro de **unidades primarias de muestreo**. Éstas quedan definidas por los: - **Estratos**: Subpoblaciones naturales que queremos estudiar (ej: sexo, edad, CCAA). - **Conglomerados**: Unidades definidas dentro de cada estrato seleccionadas por muestreo. > El resultado es que las distintas unidades no tienen la misma probabilidad de ser seleccionadas. Para obtener estimaciones **representativas** de la población debemos dar más peso a las observaciones con menor **probabilidad de inclusión**. --- ## Tipos de pesos - **Peso de diseño**: Calculado como la inversa de la probabilidad de selección. - **Peso postestratificación**: Ajustan por la tasa de no respuesta dentro de cada unidad primaria de muestreo. - **Peso normalizado**: Peso tipificado para tener media igual a 1. > Muchos procedimientos estadísticos en **R** (ej: `glm`), asumen que, si hay pesos, representan el número de observaciones. Es necesario usar *pesos normalizados* para no sobrevalorar la precisión teniendo en cuenta las distintas probabilidad de selección. > La *DHS* proporciona **pesos postestratificación normalizados** ```r DHS = DHS %>% mutate(peso=V005/1000000) DHS %>% sjmisc::descr(peso) ``` ``` ## ## ## Basic descriptive statistics ## ## var type label n NA.prc mean sd ## peso numeric Factor de ponderacion 38335 0 0.96 1.67 ## se md trimmed range iqr skew ## 0.01 0.45 0.61 28.92 (0-28.92) 0.7456075 4.7 ``` --- ## Distribución de los pesos en el ejemplo .pull-left[ Los pesos normalizados representan cuán infrarrepresentadas están las mujeres *como* la entrevistada. ```r DHS %>% ggplot(aes(y=peso,x=0),alpha=0.7) + geom_jitter(size=0.1) + geom_violin(fill="magenta") + scale_y_log10() + scale_x_discrete(labels = NULL, breaks = NULL) + labs(x = "") + theme_minimal(20) ``` > Algunas mujeres están 10 veces sobrerrepresentadas y otras 10 veces infrarrepresentadas. **Muy lejos del muestreo aleatorio**. ] .pull-right[ ![](index_files/figure-html/pesos-1.png) > **Sesgos** al no ponderar si la variable de interés se relaciona con el peso ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Tablas ponderadas <hr> **Una variable** <hr> Tabulaciones cruzadas <hr> Modelos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ Tanto `sjmisc` como `expss` y el `tidyverse` tienen funciones. .panelset[ .panel[.panel-name[Sin ponderar] ```r # install.packages("sjmisc") sjmisc::frq(DHS$V313) #Tabulación simple ``` ``` ## ## Uso actual por tipo de método (x) <numeric> ## # total N=38335 valid N=33311 mean=1.67 sd=1.39 ## ## Value | Label | N | Raw % | Valid % | Cum. % ## ------------------------------------------------------------- ## 0 | No hay método | 13017 | 33.96 | 39.08 | 39.08 ## 1 | Método folclórico | 188 | 0.49 | 0.56 | 39.64 ## 2 | Método tradicional | 4762 | 12.42 | 14.30 | 53.94 ## 3 | Método moderno | 15344 | 40.03 | 46.06 | 100.00 ## <NA> | <NA> | 5024 | 13.11 | <NA> | <NA> ``` ] .panel[.panel-name[{sjmisc}] ```r sjmisc::frq(DHS,V313,weights=peso) #Tabulación ponderada ``` ``` ## ## Uso actual por tipo de método (V313) <numeric> ## # total N=33381 valid N=33381 mean=1.44 sd=1.41 ## ## Value | Label | N | Raw % | Valid % | Cum. % ## ------------------------------------------------------------- ## 0 | No hay método | 15916 | 47.68 | 47.68 | 47.68 ## 1 | Método folclórico | 102 | 0.31 | 0.31 | 47.99 ## 2 | Método tradicional | 4237 | 12.69 | 12.69 | 60.68 ## 3 | Método moderno | 13126 | 39.32 | 39.32 | 100.00 ## <NA> | <NA> | 0 | 0.00 | <NA> | <NA> ``` > Grandes diferencias, por ejemplo en las no usuarias. ] .panel[.panel-name[{expss}] Requiere más código, pero permite controlar todo lo que se pone en las tablas, y permite pesos. ```r library(expss) DHS %>% tab_cells(V313) %>% tab_weight(peso) %>% tab_stat_cpct() %>% tab_pivot() ``` <table class='gmisc_table' style='border-collapse: collapse; margin-top: 1em; margin-bottom: 1em;' > <thead> <tr><th style='border-bottom: 1px solid grey; font-weight: 900; border-top: 2px solid grey; text-align: center;'></th> <th style='font-weight: 900; border-bottom: 1px solid grey; border-top: 2px solid grey; text-align: center;'> #Total </th> </tr> </thead> <tbody> <tr><td colspan='2' style='font-weight: 900;'> Uso actual por tipo de método </td></tr> <tr> <td style='text-align: left;'> No hay método </td> <td style='text-align: right;'>47.7</td> </tr> <tr> <td style='text-align: left;'> Método folclórico </td> <td style='text-align: right;'>0.3</td> </tr> <tr> <td style='text-align: left;'> Método tradicional </td> <td style='text-align: right;'>12.7</td> </tr> <tr> <td style='text-align: left;'> Método moderno </td> <td style='text-align: right;'>39.3</td> </tr> <tr> <td style='border-bottom: 2px solid grey; text-align: left;'> #Total cases </td> <td style='border-bottom: 2px solid grey; text-align: right;'>33311</td> </tr> </tbody> </table> ] .panel[.panel-name[{tidyverse}] ```r DHS %>% count(V313,wt=peso) %>% mutate(`% (todos)`=n/sum(n)*100, `% (validos)`=n/(sum(n)-last(n))*100) ``` ``` ## # A tibble: 5 x 4 ## V313 n `% (todos)` `% (validos)` ## * <dbl+lbl> <dbl> <dbl> <dbl> ## 1 0 [No hay método] 15916. 43.1 47.7 ## 2 1 [Método folclórico] 102. 0.277 0.306 ## 3 2 [Método tradicional] 4237. 11.5 12.7 ## 4 3 [Método moderno] 13126. 35.6 39.3 ## 5 NA 3540. 9.59 10.6 ``` ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Tablas ponderadas <hr> Una variable <hr> **Tabulaciones cruzadas** <hr> Modelos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[{sjmisc}] ```r DHS %>% group_by(V025) %>% sjmisc::frq(V313,weights=peso) # ponderada ``` ``` ## ## Uso actual por tipo de método (V313) <numeric> ## # grouped by: Urbano ## # total N=27574 valid N=27574 mean=1.42 sd=1.42 ## ## Value | Label | N | Raw % | Valid % | Cum. % ## ------------------------------------------------------------- ## 0 | No hay método | 13473 | 48.86 | 48.86 | 48.86 ## 1 | Método folclórico | 48 | 0.17 | 0.17 | 49.04 ## 2 | Método tradicional | 3130 | 11.35 | 11.35 | 60.39 ## 3 | Método moderno | 10923 | 39.61 | 39.61 | 100.00 ## <NA> | <NA> | 0 | 0.00 | <NA> | <NA> ## ## ## Uso actual por tipo de método (V313) <numeric> ## # grouped by: Rural ## # total N=5807 valid N=5807 mean=1.53 sd=1.36 ## ## Value | Label | N | Raw % | Valid % | Cum. % ## ------------------------------------------------------------ ## 0 | No hay método | 2443 | 42.07 | 42.07 | 42.07 ## 1 | Método folclórico | 54 | 0.93 | 0.93 | 43.00 ## 2 | Método tradicional | 1107 | 19.06 | 19.06 | 62.06 ## 3 | Método moderno | 2203 | 37.94 | 37.94 | 100.00 ## <NA> | <NA> | 0 | 0.00 | <NA> | <NA> ``` ] .panel[.panel-name[`flat_table`] ```r DHS %>% sjmisc::flat_table(V313,V025, margin="col",weights=peso) ``` ``` ## V025 Urbano Rural ## V313 ## No hay método 48.86 42.07 ## Método folclórico 0.17 0.93 ## Método tradicional 11.35 19.06 ## Método moderno 39.61 37.94 ``` Por defecto representa recuentos. Podemos utilizar porcentajes de fila (`"row"`), columna (`"col"`) o celda (`"cell"`). ] .panel[.panel-name[{expss}] ```r library(expss) DHS %>% tab_cells(V313) %>% tab_cols(V025) %>% tab_weight(peso) %>% tab_stat_cpct() %>% tab_pivot() ``` <table class='gmisc_table' style='border-collapse: collapse; margin-top: 1em; margin-bottom: 1em;' > <thead> <tr> <th style='border-top: 2px solid grey;'></th> <th colspan='2' style='font-weight: 900; border-bottom: 1px solid grey; border-top: 2px solid grey; text-align: center;'> Tipo de lugar de residencia </th> </tr> <tr><th style='border-bottom: 1px solid grey; font-weight: 900; text-align: center;'></th> <th style='font-weight: 900; border-bottom: 1px solid grey; text-align: center;'> Urbano </th> <th style='font-weight: 900; border-bottom: 1px solid grey; text-align: center;'> Rural </th> </tr> </thead> <tbody> <tr><td colspan='3' style='font-weight: 900;'> Uso actual por tipo de método </td></tr> <tr> <td style='text-align: left;'> No hay método </td> <td style='text-align: right;'>48.9</td> <td style='text-align: right;'>42.1</td> </tr> <tr> <td style='text-align: left;'> Método folclórico </td> <td style='text-align: right;'>0.2</td> <td style='text-align: right;'>0.9</td> </tr> <tr> <td style='text-align: left;'> Método tradicional </td> <td style='text-align: right;'>11.4</td> <td style='text-align: right;'>19.1</td> </tr> <tr> <td style='text-align: left;'> Método moderno </td> <td style='text-align: right;'>39.6</td> <td style='text-align: right;'>37.9</td> </tr> <tr> <td style='border-bottom: 2px solid grey; text-align: left;'> #Total cases </td> <td style='border-bottom: 2px solid grey; text-align: right;'>23873</td> <td style='border-bottom: 2px solid grey; text-align: right;'>9438</td> </tr> </tbody> </table> ] .panel[.panel-name[{tidyverse}] ```r DHS %>% group_by(V025) %>% count(V313,wt=peso) %>% mutate(`% (todos)`=n/sum(n)*100, `% (validos)`=n/(sum(n)-last(n))*100) %>% knitr::kable() ``` | V025| V313| n| % (todos)| % (validos)| |----:|----:|-----------:|----------:|-----------:| | 1| 0| 13473.30212| 44.6345592| 48.8618408| | 1| 1| 47.91835| 0.1587446| 0.1737791| | 1| 2| 3130.19963| 10.3697727| 11.3518805| | 1| 3| 10922.86262| 36.1854246| 39.6124995| | 1| NA| 2611.52481| 8.6514989| 9.4708712| | 2| 0| 2443.07359| 36.2678698| 42.0690825| | 2| 1| 54.31568| 0.8063260| 0.9353016| | 2| 2| 1106.50080| 16.4262047| 19.0536517| | 2| 3| 2203.39986| 32.7098700| 37.9419642| | 2| NA| 928.90274| 13.7897295| 15.9954601| ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### Tablas ponderadas <hr> Una variable <hr> Tabulaciones cruzadas <hr> **Modelos** <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[`glm`] ```r mod1=glm(I(V313==3)~V025,family=binomial, data=DHS,weights=peso) summary(mod1) ``` ``` ## ## Call: ## glm(formula = I(V313 == 3) ~ V025, family = binomial, data = DHS, ## weights = peso) ## ## Deviance Residuals: ## Min 1Q Median 3Q Max ## -4.8983 -0.7358 -0.3379 0.8611 7.3190 ## ## Coefficients: ## Estimate Std. Error z value Pr(>|z|) ## (Intercept) -0.35126 0.03657 -9.604 <2e-16 *** ## V025 -0.07037 0.02971 -2.368 0.0179 * ## --- ## Signif. codes: ## 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## (Dispersion parameter for binomial family taken to be 1) ## ## Null deviance: 44742 on 33310 degrees of freedom ## Residual deviance: 44737 on 33309 degrees of freedom ## (5024 observations deleted due to missingness) ## AIC: 40087 ## ## Number of Fisher Scoring iterations: 3 ``` ] .panel[.panel-name[`glm` con {expss}] ```r mod2=use_labels(DHS, glm(I(V313==3)~V025,family=binomial,weights=peso)) summary(mod2) ``` ``` ## ## Call: ## glm(formula = I(`Uso actual por tipo de método` == 3) ~ `Tipo de lugar de residencia V025`, ## family = binomial, weights = `Factor de ponderacion peso`) ## ## Deviance Residuals: ## Min 1Q Median 3Q Max ## -4.8983 -0.7358 -0.3379 0.8611 7.3190 ## ## Coefficients: ## Estimate Std. Error ## (Intercept) -0.35126 0.03657 ## `Tipo de lugar de residencia V025` -0.07037 0.02971 ## z value Pr(>|z|) ## (Intercept) -9.604 <2e-16 *** ## `Tipo de lugar de residencia V025` -2.368 0.0179 * ## --- ## Signif. codes: ## 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## (Dispersion parameter for binomial family taken to be 1) ## ## Null deviance: 44742 on 33310 degrees of freedom ## Residual deviance: 44737 on 33309 degrees of freedom ## (5024 observations deleted due to missingness) ## AIC: 40087 ## ## Number of Fisher Scoring iterations: 3 ``` ] .panel[.panel-name[{sjPlot}] ```r sjPlot::tab_model(mod1,dv.labels="Uso anticonceptivos modernos") ``` <table style="border-collapse:collapse; border:none;"> <tr> <th style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; text-align:left; "> </th> <th colspan="3" style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; ">Uso anticonceptivos modernos</th> </tr> <tr> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; text-align:left; ">Predictors</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">Odds Ratios</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">CI</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">p</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">(Intercept)</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.70</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.66 – 0.76</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Tipo de lugar de<br>residencia</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.93</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.88 – 0.99</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong>0.018</strong></td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm; border-top:1px solid;">Observations</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left; border-top:1px solid;" colspan="3">33311</td> </tr> </table> Produce tablas de resultados que usan etiquetas, y muy configurable. > El problema es que con encuestas complejas los errores estándar no están bien calculados: no es una muestra aleatoria. ] ] ] --- class: inverse, center, middle # Encuestas complejas: Cálculo apropiado de errores estándar. El paquete {survey} --- # ¿Por qué están mal calculados los errores estándar? - Dentro de cada **unidad primaria de muestreo** habrá menos variabilidad que en la población en general. - Si no lo tenemos en cuenta, en general estaremos **subestimando la varianza**. - El estudio de la **varianza entre grupos** me permitirá estimar la variabilidad correcta. -- ## ¿Qué hacer? - **En cada modelo**: Uso de matriz de varianzas-covarianzas `vcovCL` definiendo la variable de conglomerados. -- - **En general**: Especificar con `{survey}` la estructura de la encuesta, y usar sus funciones especializadas. --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### {survey} <hr> **Diseño de la encuesta** <hr> IC de proporciones <hr> Tabulaciones cruzadas <hr> Modelos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[Variable de interés] Hacemos las manipulaciones de variables *antes* de crear el diseño. De otro modo es más complicado (aunque podéis mirar [{srvyr}](https://cran.r-project.org/web/packages/srvyr/), y luego veremos un ejemplo de mutación directamente en un objeto `survey.design` ) NOTA: La opción `out="v" sirve para tablas con formato html. ```r DHS=DHS %>% mutate(moderno=factor((V313==3), labels=c("No","Si"))) DHS %>% sjmisc::frq(moderno,weights=peso,out="v") ``` <table style="border-collapse:collapse; border:none;"> <caption style="font-weight: bold; text-align:left;">moderno <span style="font-weight: normal; font-style: italic"><categorical></span></caption> <tr> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align:left;text-align:left; ">val</th> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align: left;">label</th> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align: right;">frq</th> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align: right;">raw.prc</th> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align: right;">valid.prc</th> <th style="border-top: double; text-align:center; font-style:italic; font-weight:normal; padding:0.2cm; border-bottom:1px solid black; text-align: right;">cum.prc</th> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left;text-align:left; ">No</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: left;"><none></td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">20255</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">60.68</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">60.68</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">60.68</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left;text-align:left; ">Si</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: left;"><none></td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">13126</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">39.32</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">39.32</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; text-align: right;">100</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left;text-align:left; border-bottom: 1px solid; ">NA</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; border-bottom: 1px solid; text-align: left;">NA</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; border-bottom: 1px solid; text-align: right;">0</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; border-bottom: 1px solid; text-align: right;">0</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; border-bottom: 1px solid; text-align: right;">NA</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; border-bottom: 1px solid; text-align: right;">NA</td> </tr> <tr> <td colspan="7" style="font-style:italic; border-top:double black; text-align:right;">total N=33381 · valid N=33381 · x̄=1.39 · σ=0.49 </td> </tr> </table> ] .panel[.panel-name[`svydesign`] Tenemos que declarar la variable con los nombres de los conglomerados, `ids`, la de estratos `strata`, y los pesos `weights`. ```r #install.packages("survey") library(survey) disDHS=svydesign(ids=~V021, strata=~V022, weights=~peso, data=DHS %>% filter(peso>0,!is.na(moderno)), nest=TRUE) summary(disDHS) ``` ``` ## Stratified 1 - level Cluster Sampling design (with replacement) ## With (3252) clusters. ## svydesign(ids = ~V021, strata = ~V022, weights = ~peso, data = DHS %>% ## filter(peso > 0, !is.na(moderno)), nest = TRUE) ## Probabilities: ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## 0.03457 0.98965 2.16821 3.36628 3.82383 94.97578 ## Stratum Sizes: ## 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 ## obs 11 42 31 33 47 80 75 125 86 710 23 95 164 117 97 ## design.PSU 2 4 4 4 5 8 8 13 9 62 4 11 15 12 10 ## actual.PSU 2 4 4 4 5 8 8 13 9 62 4 11 15 12 10 ## 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 ## obs 33 24 39 41 27 475 33 38 54 40 22 84 65 41 68 ## design.PSU 4 2 5 4 3 40 4 4 6 4 3 8 6 5 7 ## actual.PSU 4 2 5 4 3 40 4 4 6 4 3 8 6 5 7 ## 31 32 33 34 35 36 37 38 39 40 41 42 43 44 ## obs 575 100 168 237 216 154 27 49 21 62 80 123 37 79 ## design.PSU 58 11 18 23 20 15 3 6 2 6 8 12 4 8 ## actual.PSU 58 11 18 23 20 15 3 6 2 6 8 12 4 8 ## 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 ## obs 99 161 59 60 56 79 26 102 505 20 34 22 57 36 43 ## design.PSU 9 13 6 6 6 8 3 12 50 2 3 2 5 4 6 ## actual.PSU 9 13 6 6 6 8 3 12 50 2 3 2 5 4 6 ## 60 61 62 63 64 65 66 67 68 69 70 71 72 73 ## obs 29 69 32 24 799 152 248 429 201 270 23 61 77 84 ## design.PSU 4 7 3 3 66 19 28 42 17 24 4 8 9 9 ## actual.PSU 4 7 3 3 66 19 28 42 17 24 4 8 9 9 ## 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 ## obs 54 26 55 52 47 41 509 14 63 40 20 37 49 65 751 ## design.PSU 6 3 6 6 6 4 46 2 6 4 2 4 5 7 75 ## actual.PSU 6 3 6 6 6 4 46 2 6 4 2 4 5 7 75 ## 89 90 91 92 93 94 95 96 97 98 99 100 101 102 ## obs 65 64 63 74 36 42 113 37 20 779 44 65 98 68 ## design.PSU 7 7 7 7 4 5 11 4 2 64 5 8 9 7 ## actual.PSU 7 7 7 7 4 5 11 4 2 64 5 8 9 7 ## 103 104 105 106 107 108 109 110 111 112 113 114 ## obs 44 15 134 232 222 284 123 17 67 80 108 55 ## design.PSU 4 2 12 22 22 24 9 2 8 9 10 5 ## actual.PSU 4 2 12 22 22 24 9 2 8 9 10 5 ## 115 116 117 118 119 120 121 122 123 124 125 126 ## obs 115 56 26 162 102 389 52 94 136 150 95 59 ## design.PSU 13 6 3 16 10 32 7 11 14 13 9 7 ## actual.PSU 13 6 3 16 10 32 7 11 14 13 9 7 ## 127 128 129 130 131 132 133 134 135 136 137 138 ## obs 27 119 54 135 312 69 100 167 145 125 41 103 ## design.PSU 3 11 6 13 23 8 10 16 13 9 4 10 ## actual.PSU 3 11 6 13 23 8 10 16 13 9 4 10 ## 139 140 141 142 143 144 145 146 147 148 149 150 ## obs 85 167 112 225 212 388 898 982 704 133 316 301 ## design.PSU 8 16 12 16 43 49 93 92 63 15 30 28 ## actual.PSU 8 16 12 16 43 49 93 92 63 15 30 28 ## 151 152 153 154 155 156 157 158 159 160 161 162 ## obs 35 154 277 83 215 136 147 88 16 60 44 98 ## design.PSU 4 14 26 8 17 11 11 8 2 4 3 10 ## actual.PSU 4 14 26 8 17 11 11 8 2 4 3 10 ## 163 164 165 166 167 168 169 170 171 172 173 174 ## obs 76 432 102 210 248 92 62 107 287 76 78 121 ## design.PSU 8 32 12 24 25 9 7 14 26 8 9 13 ## actual.PSU 8 32 12 24 25 9 7 14 26 8 9 13 ## 175 176 177 178 179 180 181 182 183 184 185 186 ## obs 64 137 74 200 61 132 134 67 101 59 41 27 ## design.PSU 6 15 9 23 6 14 18 8 11 6 5 3 ## actual.PSU 6 15 9 23 6 14 18 8 11 6 5 3 ## 187 188 189 190 191 192 193 194 195 196 197 198 ## obs 205 94 102 339 22 44 63 95 113 81 188 212 ## design.PSU 22 11 13 34 2 4 6 8 10 8 16 18 ## actual.PSU 22 11 13 34 2 4 6 8 10 8 16 18 ## 199 200 201 202 203 204 205 206 207 208 209 210 ## obs 263 259 11 26 21 23 20 16 71 101 71 136 ## design.PSU 25 22 2 3 2 2 2 2 8 10 7 14 ## actual.PSU 25 22 2 3 2 2 2 2 8 10 7 14 ## 211 212 213 214 215 216 217 218 219 220 221 222 ## obs 492 18 96 108 81 64 50 83 59 106 191 419 ## design.PSU 53 2 9 10 6 6 5 8 6 11 19 32 ## actual.PSU 53 2 9 10 6 6 5 8 6 11 19 32 ## 223 224 225 226 228 229 230 231 232 233 234 235 ## obs 125 255 379 310 24 130 30 88 182 198 95 53 ## design.PSU 14 27 35 28 5 14 4 9 16 18 10 5 ## actual.PSU 14 27 35 28 5 14 4 9 16 18 10 5 ## 236 237 238 239 240 241 242 243 244 245 246 247 ## obs 79 140 235 90 119 35 105 201 258 341 21 57 ## design.PSU 7 14 23 9 10 4 9 19 21 29 2 6 ## actual.PSU 7 14 23 9 10 4 9 19 21 29 2 6 ## 248 249 ## obs 69 275 ## design.PSU 7 22 ## actual.PSU 7 22 ## Data variables: ## [1] "ID1" "HHID" "CASEID" ## [4] "V001" "V002" "V003" ## [7] "V004" "V007" "V008" ## [10] "V009" "V010" "V011" ## [13] "V012" "V013" "V014" ## [16] "V015" "V017" "V018" ## [19] "V019" "V019A" "V020" ## [22] "V021" "V023" "V024" ## [25] "V025" "V026" "V027" ## [28] "V028" "V029" "V030" ## [31] "V031" "V032" "V033" ## [34] "V034" "V040" "V042" ## [37] "V043" "V044" "V000" ## [40] "Q105DD" "V101" "V102" ## [43] "V103" "V104" "V105" ## [46] "V106" "V107" "V113" ## [49] "V115" "V116" "V119" ## [52] "V120" "V121" "V122" ## [55] "V123" "V124" "V125" ## [58] "V127" "V128" "V129" ## [61] "V130" "V131" "V133" ## [64] "V134" "V135" "V136" ## [67] "V137" "V138" "V139" ## [70] "V140" "V141" "V149" ## [73] "V150" "V151" "V152" ## [76] "V153" "AWFACTT" "AWFACTU" ## [79] "AWFACTR" "AWFACTE" "AWFACTW" ## [82] "V155" "V156" "V157" ## [85] "V158" "V159" "V160" ## [88] "V161" "V166" "V167" ## [91] "V168" "ML101" "QD333_1" ## [94] "QD333_2" "QD333_3" "QD333_4" ## [97] "QD333_5" "QD333_6" "UBIGEO" ## [100] "V022" "V005" "V190" ## [103] "V191" "mujeres12a49" "NCONGLOME" ## [106] "V201" "V202" "V203" ## [109] "V204" "V205" "V206" ## [112] "V207" "V208" "V209" ## [115] "V210" "V211" "V212" ## [118] "V213" "V214" "V215" ## [121] "V216" "V217" "V218" ## [124] "V219" "V220" "V221" ## [127] "V222" "V223" "V224" ## [130] "V225" "V226" "V227" ## [133] "V228" "V229" "V230" ## [136] "V231" "V232" "V233" ## [139] "V234" "V235" "V237" ## [142] "V238" "V239" "V240" ## [145] "V241" "V242" "V243" ## [148] "V301" "V302" "V310" ## [151] "V311" "V312" "V313" ## [154] "V315" "V316" "V317" ## [157] "V318" "V319" "V320" ## [160] "V321" "V322" "V323" ## [163] "V323A" "V325A" "V326" ## [166] "V327" "V337" "V359" ## [169] "V360" "V361" "V362" ## [172] "V363" "V364" "V367" ## [175] "V372" "V372A" "V375A" ## [178] "V376" "V376A" "V379" ## [181] "V380" "V384A" "V384B" ## [184] "V384C" "V393" "V394" ## [187] "V395" "V3A00A" "V3A00B" ## [190] "V3A00C" "V3A00D" "V3A00E" ## [193] "V3A00F" "V3A00G" "V3A00H" ## [196] "V3A00I" "V3A00J" "V3A00K" ## [199] "V3A00L" "V3A00M" "V3A00N" ## [202] "V3A00O" "V3A00P" "V3A00Q" ## [205] "V3A00R" "V3A00S" "V3A00T" ## [208] "V3A00U" "V3A00V" "V3A00W" ## [211] "V3A00X" "V3A00Y" "V3A00Z" ## [214] "V3A01" "V3A02" "V3A03" ## [217] "V3A04" "V3A05" "V3A06" ## [220] "V3A07" "V3A08A" "V3A08B" ## [223] "V3A08C" "V3A08D" "V3A08E" ## [226] "V3A08F" "V3A08G" "V3A08H" ## [229] "V3A08I" "V3A08J" "V3A08K" ## [232] "V3A08L" "V3A08M" "V3A08N" ## [235] "V3A08O" "V3A08P" "V3A08Q" ## [238] "V3A08R" "V3A08S" "V3A08T" ## [241] "V3A08U" "V3A08V" "V3A08W" ## [244] "V3A08X" "V3A08Z" "V3A09A" ## [247] "V3A09B" "V305_01" "V305_02" ## [250] "V305_03" "V305_05" "V305_06" ## [253] "V305_07" "V305_08" "V305_09" ## [256] "V305_10" "V305_11" "V305_13" ## [259] "V305_14" "V305_15" "V305_16" ## [262] "V307_01" "V307_02" "V307_03" ## [265] "V307_04" "V307_05" "V307_06" ## [268] "V307_07" "V307_08" "V307_09" ## [271] "V307_10" "V307_11" "V307_12" ## [274] "V307_13" "V307_14" "V307_15" ## [277] "V307_16" "QI302_05A" "QI302_05B" ## [280] "V501" "V502" "V503" ## [283] "V504" "V505" "V506" ## [286] "V507" "V508" "V509" ## [289] "V510" "V511" "V512" ## [292] "V513" "V525" "V527" ## [295] "V528" "V529" "V530" ## [298] "V531" "V532" "V535" ## [301] "V536" "V537" "V538" ## [304] "V539" "V540" "V541" ## [307] "V602" "V603" "V604" ## [310] "V605" "V613" "V614" ## [313] "V616" "V621" "V623" ## [316] "V624" "V625" "V626" ## [319] "V627" "V628" "V629" ## [322] "V631" "V632" "V633A" ## [325] "V633B" "V633C" "V633D" ## [328] "V633E" "V633F" "V633G" ## [331] "V634" "V701" "V702" ## [334] "V704" "V705" "V714" ## [337] "V714A" "V715" "V716" ## [340] "V717" "V719" "V721" ## [343] "V729" "V730" "V731" ## [346] "V732" "V739" "V740" ## [349] "V741" "V743A" "V743B" ## [352] "V743C" "V743D" "V743E" ## [355] "V743F" "V744A" "V744B" ## [358] "V744C" "V744D" "V744E" ## [361] "V746" "peso" "moderno" ``` ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### {survey} <hr> Diseño de la encuesta <hr> **Tabulaciones** <hr> Tabulaciones cruzadas <hr> Modelos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[`svymean`] - Las funciones de `survey` empiezan siempre con `svy` y exigen un objeto de encuesta con diseño. - Exigen también una fórmula, en general `~nombre`. - Al resultado se le pueden pedir otras estimaciones, por ejemplo el error estándar (`SE`) o el intervalo de confianza (`confint`) - Por defecto utiliza el intervalo al 95%. Se puede cambiar con `level`. ```r svymean(~moderno,disDHS) ``` ``` ## mean SE ## modernoNo 0.60678 0.0055 ## modernoSi 0.39322 0.0055 ``` ```r svymean(~moderno,disDHS) %>% confint() ``` ``` ## 2.5 % 97.5 % ## modernoNo 0.5959328 0.6176296 ## modernoSi 0.3823704 0.4040672 ``` ] .panel[.panel-name[`svyciprop`] - La función `svymean` sirve para variables factor o cuantitativa, pero no tiene en cuenta que se trata de una probabilidad de una binomial. - `svyciprop`, en cambio, sólo sirve para proporciones. - Se basa, por defecto, en el cálculo de un modelo logit. - En este caso, no hay cambios, pero sí los hay para probabilidades cercanas a 0 o a 1. ```r svyciprop(~moderno,disDHS) ``` ``` ## 2.5% 97.5% ## moderno 0.393 0.382 0.4 ``` ```r svyciprop(~moderno,disDHS) %>% SE() ``` ``` ## as.numeric(moderno) ## 0.005535006 ``` ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### {survey} <hr> Diseño de la encuesta <hr> Tabulaciones <hr> **Tabulaciones cruzadas** <hr> Modelos <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[`svytable`] Calcula la tabla, pero no utiliza etiquetas. ```r tabRes=svytable(~moderno+V025,disDHS) tabRes ``` ``` ## V025 ## moderno 1 2 ## No 16651.42 3603.89 ## Si 10922.86 2203.40 ``` Para utilizar etiquetas hay que convertir en factor ```r svytable(~moderno+as_factor(V025),disDHS) ``` ``` ## as_factor(V025) ## moderno Urbano Rural ## No 16651.42 3603.89 ## Si 10922.86 2203.40 ``` ] .panel[.panel-name[tests] - Se pueden calcular distintas medidas de asociación entre variables, siempre teniendo en cuenta la estructura compleja. ```r tabRes %>% summary(statistic="Chisq") ``` ``` ## V025 ## moderno 1 2 ## No 16651 3604 ## Si 10923 2203 ## ## Pearson's X^2: Rao & Scott adjustment ## ## data: svychisq(~moderno + V025, design = disDHS, statistic = "Chisq") ## X-squared = 5.5988, df = 1, p-value = 0.09966 ``` ] .panel[.panel-name[`svyby`] - Análogo al `group_by` de {tidyverse}, podemos definir las proporciones para cada nivel o los errores estándar. ```r svyby(~moderno,~V501,disDHS,svymean,vartype="ci") %>% mutate(V501=as_factor(V501)) %>% kable(digits=3) ``` | |V501 | modernoNo| modernoSi| ci_l.modernoNo| ci_l.modernoSi| ci_u.modernoNo| ci_u.modernoSi| |:--|:---------------|---------:|---------:|--------------:|--------------:|--------------:|--------------:| |0 |Nunca casada | 0.850| 0.150| 0.835| 0.135| 0.865| 0.165| |1 |Casado | 0.450| 0.550| 0.425| 0.525| 0.475| 0.575| |2 |Viviendo juntos | 0.442| 0.558| 0.426| 0.542| 0.458| 0.574| |3 |Viuda | 0.812| 0.188| 0.688| 0.064| 0.936| 0.312| |4 |Divorciada | 0.703| 0.297| 0.475| 0.070| 0.930| 0.525| |5 |No viven juntos | 0.704| 0.296| 0.675| 0.267| 0.733| 0.325| ] .panel[.panel-name[`as_tibble`] - Podemos convertir los objetos en `tibbles` que manipulamos con las funciones del `{tidyverse}` ```r svyby(~moderno,~as_factor(V501),disDHS,svymean,vartype="ci") %>% as_tibble() %>% dplyr::select(`Estado Civil`=1,ends_with("Si")) %>% kable() ``` |Estado Civil | modernoSi| ci_l.modernoSi| ci_u.modernoSi| |:---------------|---------:|--------------:|--------------:| |Nunca casada | 0.1498518| 0.1347637| 0.1649398| |Casado | 0.5500164| 0.5252821| 0.5747506| |Viviendo juntos | 0.5582293| 0.5422314| 0.5742272| |Viuda | 0.1879977| 0.0643558| 0.3116396| |Divorciada | 0.2973546| 0.0698621| 0.5248470| |No viven juntos | 0.2957780| 0.2667916| 0.3247644| ] ] ] --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) ### {survey} <hr> Diseño de la encuesta <hr> Tabulaciones <hr> Tabulaciones cruzadas <hr> **Modelos** <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[Transformar] - Vamos a estimar el modelo logit de `moderno` en función de residencia y estado civil. - Transformamos primero las variables V025 y V501 en factores. `update` es equivalente a `mutate` pero en objetos `svy`. ```r disDHS2=update(disDHS,`Estado civil`= as_factor(V501), Residencia=as_factor(V025)) ``` Paquetes útiles para transformar: {stringr}, cadenas; {forcats}, {labelled} y {sjmisc}, factores, NAs. `as_factor`, `to_factor`, `unlabelled`, `to_character` ] .panel[.panel-name[Estimar] - `svyglm` funciona igual que `glm`, pero no hace falta explicitar pesos o conglomerados al estar declarados en el diseño. ```r mod3= svyglm(moderno~`Estado civil`+Residencia, disDHS2, family=binomial()) summary(mod3) ``` ``` ## ## Call: ## svyglm(formula = moderno ~ `Estado civil` + Residencia, design = disDHS2, ## family = binomial()) ## ## Survey design: ## update(disDHS, `Estado civil` = as_factor(V501), Residencia = as_factor(V025)) ## ## Coefficients: ## Estimate Std. Error t value ## (Intercept) -1.69444 0.06116 -27.703 ## `Estado civil`Casado 1.97260 0.08028 24.570 ## `Estado civil`Viviendo juntos 2.01942 0.06941 29.094 ## `Estado civil`Viuda 0.36503 0.42723 0.854 ## `Estado civil`Divorciada 0.83638 0.56275 1.486 ## `Estado civil`No viven juntos 0.86484 0.09150 9.452 ## ResidenciaRural -0.38665 0.04603 -8.400 ## Pr(>|t|) ## (Intercept) <2e-16 *** ## `Estado civil`Casado <2e-16 *** ## `Estado civil`Viviendo juntos <2e-16 *** ## `Estado civil`Viuda 0.393 ## `Estado civil`Divorciada 0.137 ## `Estado civil`No viven juntos <2e-16 *** ## ResidenciaRural <2e-16 *** ## --- ## Signif. codes: ## 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## (Dispersion parameter for binomial family taken to be 0.9975517) ## ## Number of Fisher Scoring iterations: 4 ``` ] .panel[.panel-name[Tablas] ```r mod3 %>% sjPlot::tab_model() ``` <table style="border-collapse:collapse; border:none;"> <tr> <th style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; text-align:left; "> </th> <th colspan="3" style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; ">moderno</th> </tr> <tr> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; text-align:left; ">Predictors</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">Odds Ratios</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">CI</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">p</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">(Intercept)</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.18</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.16 – 0.21</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Estado civil actual:<br>Casado</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">7.19</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">6.14 – 8.41</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Estado civil actual:<br>Viviendo juntos</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">7.53</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">6.58 – 8.63</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Estado civil actual:<br>Viuda</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">1.44</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.62 – 3.33</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.393</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Estado civil actual:<br>Divorciada</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">2.31</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.77 – 6.95</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.137</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Estado civil actual: No<br>viven juntos</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">2.37</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">1.98 – 2.84</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">Tipo de lugar de<br>residencia: Rural</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.68</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.62 – 0.74</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong><0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm; border-top:1px solid;">Observations</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left; border-top:1px solid;" colspan="3">33311</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">R<sup>2</sup> / R<sup>2</sup> adjusted</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">0.118 / -8.796</td> </tr> </table> ] .panel[.panel-name[Plots] ```r plot3= mod3 %>% sjPlot::plot_model(show.values=TRUE) plot3+theme_bw(18) ``` ![](index_files/figure-html/unnamed-chunk-42-1.png)<!-- --> ] .panel[.panel-name[`{broom}`] ```r mod3 %>% broom::glance() ``` ``` ## # A tibble: 1 x 6 ## null.deviance df.null AIC BIC deviance df.residual ## <dbl> <int> <dbl> <dbl> <dbl> <dbl> ## 1 44648. 33310 39416. 39439. 39366. 2998 ``` ```r mod3 %>% broom::tidy() %>% knitr::kable(digits=3) ``` |term | estimate| std.error| statistic| p.value| |:-----------------------------|--------:|---------:|---------:|-------:| |(Intercept) | -1.694| 0.061| -27.703| 0.000| |`Estado civil`Casado | 1.973| 0.080| 24.570| 0.000| |`Estado civil`Viviendo juntos | 2.019| 0.069| 29.094| 0.000| |`Estado civil`Viuda | 0.365| 0.427| 0.854| 0.393| |`Estado civil`Divorciada | 0.836| 0.563| 1.486| 0.137| |`Estado civil`No viven juntos | 0.865| 0.092| 9.452| 0.000| |ResidenciaRural | -0.387| 0.046| -8.400| 0.000| ] ] ] --- class: inverse, center, middle # ¿Y qué pasa si no tengo en cuenta el diseño complejo? El efecto del diseño. --- ## Subestimación de la variabilidad - Si no tenemos en cuenta el **diseño complejo** subestimamos la variabilidad de nuestros estimadores. - Pensaremos que nuestros *estimadores* son más precisos de lo que son, los *intervalos de confianza* serán demasiado estrechos, y los *contrastes* producirán p-valores demasiado pequeños. - Al cociente entre la varianza muestral *correcta* y la que asume *muestreo aleatorio* se le denomina **efecto del diseño** - El paquete `{survey}` permite calcularlo directamente en `svymean`. También lo podemos calcular especificando diseños alternativos. - Valores típicos ~4 (doble `\(SE\)`). --- .left-column[ ##Ejemplo: Uso de anticonceptivos en la [**ENDES-DHS Perú 2019**](http://iinei.inei.gob.pe/microdatos/index.htm) <hr> ### Efecto del diseño <hr> .footnote[Más extenso [aquí](https://rpubs.com/jaortega/EncuestaR2)] ] .right-column[ .panelset[ .panel[.panel-name[Directo] ```r svymean(~moderno,disDHS2,deff="replace") ``` ``` ## mean SE DEff ## modernoNo 0.606781 0.005535 4.2771 ## modernoSi 0.393219 0.005535 4.2771 ``` > Los errores estándar son **más del doble** de los que se calcularían bajo muestreo aleatorio, la varianza más del cuádruple. ] .panel[.panel-name[Diseños alternativos] ```r # Diseño sin uso de pesos DHSunw=svydesign(ids=~1,strata=NULL,weights=NULL, data=DHS %>% filter(peso>0,!is.na(moderno))) # Diseño con pesos pero sin estructura compleja DHSw=svydesign(ids=~1,strata=NULL,weights=~peso, data=DHS %>% filter(peso>0,!is.na(moderno))) ``` ] .panel[.panel-name[Cálculo] ```r SE_comp=svyciprop(~moderno,disDHS) %>% SE() SE_w= svyciprop(~moderno,DHSw) %>% SE() SE_unw= svyciprop(~moderno,DHSunw) %>% SE() tribble(~Método,~SE, "Encuesta Compleja",SE_comp, "Con pesos",SE_w, "Sin pesos",SE_unw) %>% mutate(var=SE^2,deff=var/last(var)) %>% kable(digits=6) ``` |Método | SE| var| deff| |:-----------------|--------:|-------:|--------:| |Encuesta Compleja | 0.005535| 3.1e-05| 4.107447| |Con pesos | 0.005215| 2.7e-05| 3.646239| |Sin pesos | 0.002731| 7.0e-06| 1.000000| - La parte más grande del efecto del diseño se capta utilizando pesos. - En [EncuestaR2](https://rpubs.com/jaortega/EncuestaR2) también se presenta cómo usar `vcovCL`. ] ] ] --- class: inverse, center, middle # Conclusiones: Buenas prácticas con datos de encuesta en R --- ## Buenas prácticas con datos de encuesta. #### **IMPORTAR**: > Con {haven}. Si da [problemas](https://rpubs.com/jaortega/EncuestaR1b), {memisc} #### **TABLAS**: > Con {sjmisc}, fácil, o {expss}, potente. Con pesos. ### MODELOS > Transformar primero variables a factores o numéricas #### **MODELOS ESTÁNDAR**: > Declarar diseño y estimar con {survey}. Tablas y gráficos con {sjPlot} #### **MODELOS NO DISPONIBLES EN SURVEY**: > Con el paquete correspondiente. Usar **pesos** y varianzas cluster. --- ## Referencias ### Ejemplos más extensos, con referencias abundantes: Ortega, J. A. (2020a) Lectura de encuestas INE y análisis microeconométrico, EncuestaR - Análisis de Encuestas con R. Curso del Doctorado en Economía 2020-21. Universidad de Salamanca. <https://rpubs.com/jaortega/EncuestaR1> Ortega, J. A. (2020b) Análisis de encuestas complejas con survey. EncuestaR - Análisis de Encuestas con R. Curso del Doctorado en Economía 2020-21. Universidad de Salamanca. <https://rpubs.com/jaortega/EncuestaR2> ### Ejemplos de código para encuestas internacionales: Damico, A. J. (2020) “Analyze survey data for free”, <http://asdfree.com/> --- class: inverse, center, middle ![](assets/img/usal800.jpg) # ¡Gracias! ![](assets/img/viernesdemarzo3.png)