PIAAC data are made openly available for research purposes. They provide unique opportunities to analyse adults’ cognitive skills needed in the labour market and society, together with a rich collection of background variables dealing with educational background, labour market status, use of skills at work and outside work, among others.
PIAAC, like several other international large-scale assessments of skills, employ complex survey designs (which often involve stratified multi-stage sampling) in collecting data and complicated methodology in measuring cognitive skills of the respondents. Thus, valid statistical data analyses of PIAAC data require methodology which respects the characteristics of the survey. First, the sampling design used in collecting the data must be considered, as the appropriate approach to estimation of variances and standard errors depends on it. Second, the cognitive skills are assessed using an approach based on so-called plausible values. A valid analysis of plausible values data requires specific approach, which has its roots in the methodology of multiple imputation and analysing multiple-imputed data.
This document aims to serve as an introduction to the key issues mentioned below:
- What kind of survey designs have been adopted in PIAAC, what kind of methodology is available for analyses which successfully take the design into account, and what are the consequences if the design is not respected in the analyses.
- What are the plausible values, why they have become the standard approach in the international large-scale assessments of skills, and how valid analyses of plausible values data are conducted.
In addition, references to available technical tools for valid analysis of PIAAC data are given.
