Corrections to the PISA data for Austria
The OECD document Education Working Paper No. 5 PISA 2000: Sample Weight Problems in Austria available from the OECD Education Web Site at workingpapers describes some modification to the PISA 2000 data to reproduce the Austrian PISA target population more faithfully.
The book
describes these modifications to the Austrian PISA 2000 data also; furthermore it presents some additional (minor) modifications to the Austrian and international data for PISA 2003. As described in the book (and also discussed in some OECD publications), there were some problems with imputed data for PISA 2003. The PISA methodology creates domain specific scores for all domains and all students. For example, there are reading scores even for students who did not do any reading test assignments. These scores are called imputed scores, and in some countries these scores are biased. We supply data correcting these biases, allowing to compute modified country scores for the minor PISA 2003 domains reading, science, and problem solving. In Austria, there is an additional bias regarding stratification of the student population: Some strata are over- or underrepresented in the sample when compared with the official school statistics. Combining the imputation bias and the small misrepresentation of the school population for Austria produces modified results almost identical to the official results from the OECD PISA reports. Therefore, the official Austrian country scores for PISA 2003 may be used without intruding biases. For more detailed studies for Austria, however, and to investigate the influence of the imputation bias on country results, we suggest to use our modifications to the imputed student scores. We offer corrections to these scores for all countries.
From this page, you can download data files with these corrections.
The files contain some modified PISA variables and some additional variables.
There are 4 different files:
- Austrian corrections for PISA 2000 (CSV file compressed with ZIP)
- International corrections for PISA 2003 (CSV file compressed with ZIP)
- Austrian corrections for PISA 2003 (CSV file compressed with ZIP)
- Austrian corrections combined for PISA 2000 and PISA 2003 (MS Excel file compressed with ZIP)
The Austrian corrections for PISA 2000 give modified weights.
The file contains the original weight variables
CASEWTREAD, CASEWTMATH, CASEWTSCIE
and new case weight variables
CASEWTREADautcorr, CASEWTMATHautcorr, CASEWTSCIEautcorr
Additionally, the variable STRATUM gives the schooltype as documented in pisa20002003austriancorrections.xls.
STRATUMpure gives the schooltypes for PISA 2003 in Austria.
The PISA 2003 data set contains a variable STRATUM, and this variable encodes schooltypes, but there are two additional categories, small schools and very small schools, and the presence of these categories inhibit computations for schooltypes in Austria.
STRATUMpure just gives the original STRATUM value for small and very small schools, otherwise it is identical to STRATUM.
The international corrections for PISA 2003 need 3 new variables
readmod, sciemod, probmod
These variables have to be added to the original plausible value variables
readmod to PV1READ to PV5READ
sciemod to PV1SCIE to PV5SCIE
probmod to PV1PROB to PV5PROB
to produce variables with the corrections described in the report mentioned above.
Additionally, all the correction files for 2003 contain a variable
CASEWTautcorr
containing modified case weights for the Austrian students.