Rrepest

A way to run estimations with weighted replicate samples and plausible values

Table of Contents

• Description
• Installation
• Documentation
• Examples of use cases
• Authors
• Contributing

Description

It estimates statistics using replicate weights (Balanced Repeated Replication (BRR) weights, Jackknife replicate weights,...), thus accounting for complex survey designs in the estimation of sampling variances. It is designed specifically for use with the international education datasets produced by the OECD (e.g. PIAAC, PISA, SSES, TALIS, etc.), but works for any educational large-scale assessment and survey that uses replicated weights (e.g. ICCS, ICILS, PIRLS, TIMSS - all produced by IEA). It also allows for analyses with multiply imputed variables (plausible values); where plausible values are used, the average estimator across plausible values is reported and the imputation error is added to the variance estimator.

Installation

Using CRAN (latest official version)

Run the following code:

install.packages("Rrepest")

Using tar.gz file (latest development version)

Download Rrepest, then run

Run the following code replacing "You_R_Name" with your username:

install.packages("C:/Users/You_R_Name/Downloads/Rrepest.tar.gz",
repos = NULL,
type ="source")

Run:

library(Rrepest)

Using a GitLab token (latest development version)

Run the following code replacing "MY_TOKEN" with your gitlab token:

remotes::install_gitlab("edu_data/rrepest", host = "https://algobank.oecd.org:4430", upgrade = "never", auth_token = "MY_TOKEN")

Note: It will take a few minutes to install.

Run:

library(Rrepest)

Note: Ensure you have the package data.table installed. For a complete list of the dependencies used, consult the Description file.

Documentation

• Full documentation of Rrepest is available here.
• Cheat sheet including an overview of the syntax and auxiliaries of Rrepest is available here.
• Information on how to incorporate analyses that are not pre-programmed into Rrepest is available in the following wiki.

Examples of use cases

Rrepest supports summary statistics (i.e. mean, variance, standard deviation, quantiles, inter-quantile range), frequency count, correlation, linear regression and any other statistics that are not pre-programmed into Rrepest but take a data frame and weights as parameters (see General analysis below). Rrepest also has optional features that provide means, among others, to specify the level of analysis, obtain estimates for each level of a given categorical variable, test for differences, flag estimates that are based on fewer observations than required for reporting, compute averages. More detail on the optional features of Rrepest can be found here.

Summary statistics

# PISA 2018 Data
# df.qqq <- readRDS("//oecdmain/asgenedu/EDUCATION_DATALAKE/sources/PISA/PISA 2018/R/STU/CY07_MSU_STU_QQQ.rds")

Rrepest::Rrepest(data = df.qqq,
        svy = "PISA2015",
        est = est(c("mean","var","std","quant",0.5,"iqr",c(.9,.1)),"age"),
        by = c("cnt"))

Frequency count

# TALIS 2018 Data
# df.t <- readRDS("//oecdmain/asgenedu/EDUCATION_DATALAKE/sources/TALIS/2018/R/International/TTGINTT3.rds")

Rrepest::Rrepest(data = df.t,
                 svy = "TALISTCH",
                 est = est("freq","tt3g01"),
                 by = "cntry")

Correlation

# PISA 2018 Data
# df.qqq <- readRDS("//oecdmain/asgenedu/EDUCATION_DATALAKE/sources/PISA/PISA 2018/R/STU/CY07_MSU_STU_QQQ.rds")

Rrepest::Rrepest(data = df.qqq,
        svy = "PISA2015",
        est = est("corr",c("pv@math","pv@read")),
        by = c("cnt"))

Linear regression

# TALIS 2018 Data
# df.t <- readRDS("//oecdmain/asgenedu/EDUCATION_DATALAKE/sources/TALIS/2018/R/International/TTGINTT3.rds")

df.t <- df.t %>% 
        mutate(TT3G01_rec = case_when(TT3G01 == 2 ~ 1,
                                      TT3G01 == 1 ~ 0))

Rrepest::Rrepest(data = df.t,
        svy = "TALISTCH",
        est = est("lm","tt3g01_rec","tt3g39c"),
        by = "cntry")

Further examples can be found in the Examples.R file.

General analysis

To incorporate analyses that are not pre-programmed into Rrepest, you can utilize the 'gen' option within the est() function of Rrepest. Any line of code that takes a data frame and weights as parameters can be used with the 'gen' option. For more information, please see the following wiki.

Authors

Francesco Avvisati, Rodolfo Ilizaliturri and François Keslair.

Contact us if you want to join!

Contributing

Do you have suggestions or comments? Please open an issue.