vacs.Rmd

# Violence Against Children And Youth Surveys (VACS) {-}

[![License: GPL v3](https://img.shields.io/badge/License-GPLv3-blue.svg)](https://www.gnu.org/licenses/gpl-3.0) <img src='https://img.shields.io/badge/Tested%20Locally-Windows%20Laptop-brightgreen' alt='Local Testing Badge'>

The global surveillance system to track and monitor the burden of violence against children.

* One table per country with one row per sampled respondent.

* Nationally representative cross-sectional household surveys of children and youth ages 13–24.

* No listed update frequency across the participating nations.

* Led by the [CDC](https://www.cdc.gov/) through funding from [PEPFAR](https://www.state.gov/pepfar/), in partnership with [Together for Girls](https://www.togetherforgirls.org/en/partners).

---

## Recommended Reading {-}

Four Example Strengths & Limitations:

✔️ [Split sample interviews male and female participants in different communities to eliminate risk of harm](https://www.cdc.gov/violence-against-children/php/country-process/index.html)

✔️ [Some PEPFAR recipient nations coordinate voluntary HIV testing with face-to-face interviews](https://doi.org/10.1542/peds.2022-057124)

❌ [Must balance standardization of procedures and protocols and adaptation for culture and context](https://www.cdc.gov/violence-against-children/media/pdfs/VACS-trainingwhitepaper.pdf)

❌ [Certain sampled households excluded due to cyclones or armed attacks carried out by terrorists](https://www.togetherforgirls.org/en/resources/mozambique-vacs-report-2022)

<br>

Three Example Findings:

1. [Kenya observed significant declines in the prevalence of lifetime violence and some risk factors for violence and HIV, and improvements in some service seeking indicators between 2010 and 2019](https://doi.org/10.1016/S2214-109X(21)00457-5).

2. [In 2015, 49% of 13-24 year old Ugandan females reported any lifetime history of sexual violence](https://doi.org/10.1371/journal.pone.0260986).

3. [Over 2015-2016, 22% of Rwandan males age 13-17 reported physical violence in the past year](https://doi.org/10.1186/s12889-022-14815-0).

<br>

Two Methodology Documents:

> [Sampling design and methodology of the Violence Against Children and Youth Surveys](https://www.ncbi.nlm.nih.gov/pubmed/30472679)

> [Violence Against Children Surveys (VACS): Towards a global surveillance system](http://injuryprevention.bmj.com/content/22/Suppl_1/i17.long)

<br>

One Haiku:

```{r}
# enable us to
# lodge cane between each spoke of
# cycles of abuse
```

---

## Download, Import, Preparation {-}

1. Request public VACS data at https://www.togetherforgirls.org/en/analyzing-public-vacs-data.

2. Select the Mozambique 2019 dataset and Stata option.

3. Download and unzip the `Mozambique VACS Public Use Dataset` files:

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

vacs_tbl <- read_stata( file.path( path.expand( "~" ) , "mozambique_public use data.dta" ) )

vacs_df <- data.frame( vacs_tbl )

names( vacs_df ) <- tolower( names( vacs_df ) )
```

### Save Locally \ {-}

Save the object at any point:

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

Load the same object:

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

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

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

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

vacs_design <- 
	svydesign(
		ids = ~cluster ,
		strata = ~strata ,
		weights = ~sampleweight ,
		data = subset( vacs_df , sampleweight > 0 ) ,
		nest = TRUE
	)
```

### Variable Recoding {-}

Add new columns to the data set:
```{r eval = FALSE , results = "hide" }
vacs_design <- 
	update( 
		vacs_design , 
		
		one = 1 ,
				
		age_sex_group =
			factor( 
				ifelse( agegrp == 1 , sex , sex + 2 ) , 
				levels = 1:4 , 
				labels = c( 'male 13-17' , 'female 13-17' , 'male 18-24' , 'female 18-24' ) 
			) ,
		
		sex = factor( sex , levels = 1:2 , labels = c( 'male' , 'female' ) ) ,
		
		agegrp = factor( agegrp , levels = 1:2 , labels = c( '13-17' , '18-24' ) ) ,
		
		ever_attended_school = ifelse( eversch %in% 1:2 , as.numeric( eversch == 1 ) , NA ) ,
		
		childhood_physical_violence = as.numeric( pv18 == 1 ) ,
		
		marry =
			factor(
				marry ,
				levels = 1:3 ,
				labels = 
					c( 'Yes, ever married' , 'Yes, ever lived with a partner' , 
					'No, never married or lived with a partner' )

			) ,
			
		age_at_first_pregnancy = ifelse( prage < 98 , prage , NA ) ,
			
		age_at_first_cohabitation = ifelse( marage < 98 , marage , NA )
	)
```

---

## 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( vacs_design , "sampling" ) != 0 )

svyby( ~ one , ~ age_sex_group , vacs_design , unwtd.count )
```

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

svyby( ~ one , ~ age_sex_group , vacs_design , svytotal )
```

### Descriptive Statistics {-}

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

svyby( ~ age_at_first_cohabitation , ~ age_sex_group , vacs_design , svymean , na.rm = TRUE )
```

Calculate the distribution of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svymean( ~ marry , vacs_design )

svyby( ~ marry , ~ age_sex_group , vacs_design , svymean )
```

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

svyby( ~ age_at_first_cohabitation , ~ age_sex_group , vacs_design , svytotal , na.rm = TRUE )
```

Calculate the weighted sum of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ marry , vacs_design )

svyby( ~ marry , ~ age_sex_group , vacs_design , svytotal )
```

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

svyby( 
	~ age_at_first_cohabitation , 
	~ age_sex_group , 
	vacs_design , 
	svyquantile , 
	0.5 ,
	ci = TRUE , na.rm = TRUE
)
```

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

### Subsetting {-}

Restrict the survey design to :
```{r eval = FALSE , results = "hide" }
sub_vacs_design <- subset( vacs_design , childhood_physical_violence == 1 )
```
Calculate the mean (average) of this subset:
```{r eval = FALSE , results = "hide" }
svymean( ~ age_at_first_cohabitation , sub_vacs_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_at_first_cohabitation , vacs_design , na.rm = TRUE )

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

grouped_result <-
	svyby( 
		~ age_at_first_cohabitation , 
		~ age_sex_group , 
		vacs_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( vacs_design )
```

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

# SRS with replacement
svymean( ~ age_at_first_cohabitation , vacs_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( ~ ever_attended_school , vacs_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_at_first_cohabitation ~ ever_attended_school , vacs_design )
```

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

Perform a survey-weighted generalized linear model:
```{r eval = FALSE , results = "hide" }
glm_result <- 
	svyglm( 
		age_at_first_cohabitation ~ ever_attended_school + marry , 
		vacs_design 
	)

summary( glm_result )
```

---

## Replication Example {-}

This example matches statistics and confidence intervals within 0.1% from the Final Report of the [Mozambique Violence Against Children and Youth Survey (VACS), 2019](https://files.mutualcdn.com/tfg/assets/files/Mozambique-VACS-report.pdf), `Table 4.1.1. Prevalence of different types of sexual violence[1] before age 18, among 18-24-year-olds`:

```{r eval = FALSE , results = "hide" }
females_18_to_24_design <- subset( vacs_design , sex == 'female' & agegrp == '18-24' )

# define a function to check unweighted N, prevalence, confidence interval for each estimate
check_sv <- 
	function( this_variable , this_design = females_18_to_24_design , N , prevalence , lb , ub ){
	
		this_formula <- as.formula( paste( "~ as.numeric(" , this_variable , "== 1 )" ) )
	
		stopifnot( coef( unwtd.count( this_formula , this_design ) ) == N )
		
		this_result <- svymean( this_formula , this_design , na.rm = TRUE )
		
		stopifnot( round( coef( this_result ) , 3 ) == prevalence )
		
		stopifnot( abs( confint( this_result )[1] - lb ) < 0.0015 )
		
		stopifnot( abs( confint( this_result )[2] - ub ) < 0.0015 )
		
		invisible( TRUE )
	}
	
# sexual touching in childhood
check_sv( "sv1_only18" , N = 1232 , prevalence = 0.066 , lb = 0.039 , ub = 0.093 )

# unwanted attempted sex in childhood
check_sv( "sv2_only18" , N = 1232 , prevalence = 0.061 , lb = 0.035 , ub = 0.087 )

# pressured or coerced sex in childhood
check_sv( "sv4_only18" , N = 1221 , prevalence = 0.056 , lb = 0.035 , ub = 0.077 )

# physically forced sex in childhood
check_sv( "sv3_only18" , N = 1231 , prevalence = 0.035 , lb = 0.020 , ub = 0.051 )
```

---

## 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 VACS users, this code replicates previously-presented examples:

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

vacs_srvyr_design %>%
	group_by( age_sex_group ) %>%
	summarize( mean = survey_mean( age_at_first_cohabitation , na.rm = TRUE ) )
```