gss.Rmd

# General Social Survey (GSS) {-}

[![License: GPL v3](https://img.shields.io/badge/License-GPLv3-blue.svg)](https://www.gnu.org/licenses/gpl-3.0) <a href="https://github.com/asdfree/gss/actions"><img src="https://github.com/asdfree/gss/actions/workflows/r.yml/badge.svg" alt="Github Actions Badge"></a> 

A historical record of the concerns, experiences, attitudes, and practices of residents of the United States. 

* Both cross-sectional and panel tables with one row per sampled respondent.

* A complex sample survey generalizing to non-institutionalized adults (18+) in the United States.

* Updated biennially since 1972.

* Funded by [National Science Foundation](http://www.nsf.gov/), administered by the [National Opinion Research Center](http://www.norc.org/).

---

## Recommended Reading {-}

Four Example Strengths & Limitations:

✔️ [Fifty years of comparable measures for trend analyses](https://www.annualreviews.org/content/journals/10.1146/annurev-soc-121919-054838)

✔️ [Fifteen minutes of questionnaire aligns with International Social Survey Programme](https://new.nsf.gov/funding/opportunities/general-social-survey-competition/nsf24-537/solicitation#pgm_desc_txt)

❌ [One adult per household interviewed, living in larger households lowers probability of selection](https://www.norc.org/content/dam/norc-org/documents/standard-projects-pdf/Research%20Highlights-GSS%20(ECON)%20FINAL%20slightly%20different%20layout-DTP%20Formatted.pdf)

❌ [In 2022, 21% of mixed-mode interviews were aged 65+ versus 10% of the web-only oversample](https://gss.norc.org/content/dam/gss/get-documentation/pdf/reports/methodological-reports/GSS MR138 AmeriSpeak Oversample.pdf#page=5)

<br>

Three Example Findings:

1. [Between 2000 and 2021, confidence in the scientific community has remained steady](https://www.c-span.org/classroom/document/?18739).

2. [During 2018 to 2021, support for gun control was overwhelmingly positive among survey respondents, differing from sentiment analysis of social media data over the same period](https://gss.norc.org/content/dam/gss/get-documentation/pdf/reports/project-reports/GSS%20Project%20Report%2034%20GSMA.pdf).

3. [In 2021, 24% of Americans reported they were "not too happy" in life, up from 13% in 2018](https://www.wsj.com/articles/still-feeling-pandemic-miserable-there-are-ways-to-dig-out-11643725802).

<br>

Two Methodology Documents:

> [DOCUMENTATION AND PUBLIC USE FILE CODEBOOK (Release 1)](https://gss.norc.org/Documents/codebook/GSS%202022%20Codebook.pdf)

> [Wikipedia Entry](https://en.wikipedia.org/wiki/General_Social_Survey)

<br>

One Haiku:

```{r}
# chat about who will
# be allowed marriage, children.
# first date questionnaire
```

---

## Download, Import, Preparation {-}

Download and import the 1972-2022 cumulative data file:
```{r eval = FALSE , results = "hide" }
library(haven)

zip_tf <- tempfile()

zip_url <- "https://gss.norc.org/Documents/sas/GSS_sas.zip"
	
download.file( zip_url , zip_tf , mode = 'wb' )

unzipped_files <- unzip( zip_tf , exdir = tempdir() )

gss_tbl <- read_sas( grep( '\\.sas7bdat$' , unzipped_files , value = TRUE ) )

gss_df <- data.frame( gss_tbl )

names( gss_df ) <- tolower( names( gss_df ) )

gss_df[ , 'one' ] <- 1
```

### Save Locally \ {-}

Save the object at any point:

```{r eval = FALSE , results = "hide" }
# gss_fn <- file.path( path.expand( "~" ) , "GSS" , "this_file.rds" )
# saveRDS( gss_df , file = gss_fn , compress = FALSE )
```

Load the same object:

```{r eval = FALSE , results = "hide" }
# gss_df <- readRDS( gss_fn )
```

### Survey Design Definition {-}
Construct a complex sample survey design:

```{r eval = FALSE , results = "hide" }
library(survey)

options( survey.lonely.psu = "adjust" )

gss_design <- 
	svydesign( 
		~ vpsu , 
		strata = ~ interaction( year , vstrat ) , 
		data = subset( gss_df , year >= 1975 & !is.na( wtssnrps ) ) , 
		weights = ~ wtssnrps , 
		nest = TRUE 
	)
```

### Variable Recoding {-}

Add new columns to the data set:
```{r eval = FALSE , results = "hide" }
gss_design <- 
	update( 
		gss_design , 

		polviews = 
			factor( polviews , levels = 1:7 ,
				labels = c( "Extremely liberal" , "Liberal" ,
				"Slightly liberal" , "Moderate, middle of the road" ,
				"Slightly conservative" , "Conservative" ,
				"Extremely conservative" )
			) ,
		
		born_in_usa = as.numeric( born == 1 ) ,
		
		race = factor( race , levels = 1:3 , labels = c( "white" , "black" , "other" ) ) ,
		
		region = 
			factor( region , levels = 1:9 ,
				labels = c( "New England" , "Middle Atlantic" ,
					"East North Central" , "West North Central" ,
					"South Atlantic" , "East South Central" ,
					"West South Central" , "Mountain" , "Pacific" )
			)

	)
```

---

## Analysis Examples with the `survey` library \ {-}

### Unweighted Counts {-}

Count the unweighted number of records in the survey sample, overall and by groups:
```{r eval = FALSE , results = "hide" }
sum( weights( gss_design , "sampling" ) != 0 )

svyby( ~ one , ~ region , gss_design , unwtd.count )
```

### Weighted Counts {-}
Count the weighted size of the generalizable population, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ one , gss_design )

svyby( ~ one , ~ region , gss_design , svytotal )
```

### Descriptive Statistics {-}

Calculate the mean (average) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svymean( ~ age , gss_design , na.rm = TRUE )

svyby( ~ age , ~ region , gss_design , svymean , na.rm = TRUE )
```

Calculate the distribution of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svymean( ~ race , gss_design , na.rm = TRUE )

svyby( ~ race , ~ region , gss_design , svymean , na.rm = TRUE )
```

Calculate the sum of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ age , gss_design , na.rm = TRUE )

svyby( ~ age , ~ region , gss_design , svytotal , na.rm = TRUE )
```

Calculate the weighted sum of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ race , gss_design , na.rm = TRUE )

svyby( ~ race , ~ region , gss_design , svytotal , na.rm = TRUE )
```

Calculate the median (50th percentile) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svyquantile( ~ age , gss_design , 0.5 , na.rm = TRUE )

svyby( 
	~ age , 
	~ region , 
	gss_design , 
	svyquantile , 
	0.5 ,
	ci = TRUE , na.rm = TRUE
)
```

Estimate a ratio:
```{r eval = FALSE , results = "hide" }
svyratio( 
	numerator = ~ adults , 
	denominator = ~ hompop , 
	gss_design ,
	na.rm = TRUE
)
```

### Subsetting {-}

Restrict the survey design to females:
```{r eval = FALSE , results = "hide" }
sub_gss_design <- subset( gss_design , sex == 2 )
```
Calculate the mean (average) of this subset:
```{r eval = FALSE , results = "hide" }
svymean( ~ age , sub_gss_design , na.rm = TRUE )
```

### Measures of Uncertainty {-}

Extract the coefficient, standard error, confidence interval, and coefficient of variation from any descriptive statistics function result, overall and by groups:
```{r eval = FALSE , results = "hide" }
this_result <- svymean( ~ age , gss_design , na.rm = TRUE )

coef( this_result )
SE( this_result )
confint( this_result )
cv( this_result )

grouped_result <-
	svyby( 
		~ age , 
		~ region , 
		gss_design , 
		svymean ,
		na.rm = TRUE 
	)
	
coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )
```

Calculate the degrees of freedom of any survey design object:
```{r eval = FALSE , results = "hide" }
degf( gss_design )
```

Calculate the complex sample survey-adjusted variance of any statistic:
```{r eval = FALSE , results = "hide" }
svyvar( ~ age , gss_design , na.rm = TRUE )
```

Include the complex sample design effect in the result for a specific statistic:
```{r eval = FALSE , results = "hide" }
# SRS without replacement
svymean( ~ age , gss_design , na.rm = TRUE , deff = TRUE )

# SRS with replacement
svymean( ~ age , gss_design , na.rm = TRUE , deff = "replace" )
```

Compute confidence intervals for proportions using methods that may be more accurate near 0 and 1. See `?svyciprop` for alternatives:
```{r eval = FALSE , results = "hide" }
svyciprop( ~ born_in_usa , gss_design ,
	method = "likelihood" , na.rm = TRUE )
```

### Regression Models and Tests of Association {-}

Perform a design-based t-test:
```{r eval = FALSE , results = "hide" }
svyttest( age ~ born_in_usa , gss_design )
```

Perform a chi-squared test of association for survey data:
```{r eval = FALSE , results = "hide" }
svychisq( 
	~ born_in_usa + race , 
	gss_design 
)
```

Perform a survey-weighted generalized linear model:
```{r eval = FALSE , results = "hide" }
glm_result <- 
	svyglm( 
		age ~ born_in_usa + race , 
		gss_design 
	)

summary( glm_result )
```

---

## Replication Example {-}

Match the unweighted record count totals on [PDF page 74](https://gss.norc.org/content/dam/gss/get-documentation/pdf/codebook/GSS%202021%20Codebook.pdf#page=74) of the Public Use File codebook:
```{r eval = FALSE , results = "hide" }
stopifnot( nrow( subset( gss_design , year == 2021 ) ) == 4032 )
```

---

## Analysis Examples with `srvyr` \ {-}

The R `srvyr` library calculates summary statistics from survey data, such as the mean, total or quantile using [dplyr](https://github.com/tidyverse/dplyr/)-like syntax. [srvyr](https://github.com/gergness/srvyr) allows for the use of many verbs, such as `summarize`, `group_by`, and `mutate`, the convenience of pipe-able functions, the `tidyverse` style of non-standard evaluation and more consistent return types than the `survey` package. [This vignette](https://cran.r-project.org/web/packages/srvyr/vignettes/srvyr-vs-survey.html) details the available features. As a starting point for GSS users, this code replicates previously-presented examples:

```{r eval = FALSE , results = "hide" }
library(srvyr)
gss_srvyr_design <- as_survey( gss_design )
```
Calculate the mean (average) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
gss_srvyr_design %>%
	summarize( mean = survey_mean( age , na.rm = TRUE ) )

gss_srvyr_design %>%
	group_by( region ) %>%
	summarize( mean = survey_mean( age , na.rm = TRUE ) )
```