Harmonized variable names

February 17, 2023 · View on GitHub

back to the README :leftwards_arrow_with_hook:

This page list the harmonized variable names that will be used in all GLADs regardless of which assessment, year or country the data comes from. The first table has the variables that are included in all data sets. The second table includes variable names we have harmonized but exist only in some data sets.

varnamevarclassvarlabelvartypenote
surveyidkeySurveyID (Region_Year_Assessment)String 
countrycodekeyWB country code (3 letters)String(a) 
national_levelkeyIdcntry_raw is a national levelIndicator (1=National)(a)
idcntry_rawidCountry ID, as coded in rawdataNumerical or String(b)
idschoolidSchool IDNumerical
idgradeidGrade IDNumerical
idclassidClass IDNumerical(c)
idlearneridLearner IDNumerical
score_[assessment]_[subject]_[pv]value[Plausible value pv:] assessment score for subjectNumerical
level_[assessment]_[subject]_[pv]value[Plausible value pv:] assessment level for subjectCategorical
agetraitLearner age at time of assessmentNumerical 
urbantraitSchool is located in urban/rural areaIndicator (1=Urban)
maletraitLearner gender is male/femaleIndicator (1=Male) 
escstraitLearner socio-economic status (Purposefully not labeled yet)Numerical
learner_weightsampleTotal learner weightNumerical 

Notes:

For all assessment-years, the id variables (idcntry_raw, idschool, idgrade, idclass, idlearner) compose a unique id.

(a) The full correspondence of countrycode, national_level and idcntry_raw is found in the master countrycode list. Some examples:

  • in LLECE 1997 the countrycode MEX is linked to both the sample from the country Mexico (idcntry_raw = 21) and for the sample from the subnational unit of Nueva Leon (idntry_raw = 11). However, the first is considered national_level of 1, while the later is national_level of 0. That means that both samples are found in the GLAD module ALL, but the module CLO for Mexico is calculated using only the first sample, discarding the later.
  • in PIRLS 2001 the countrycode GBR is linked to both the samples from England (idcntry_raw = 926) and Scotland (idcntry_raw = 927) and both are considered national_level of 1. That means that both samples are found in the GLAD module ALL and the module CLO for United Kingdom is calculated pooling both samples without distinction.

(b) The variable idcntry_raw is preserved as found in the raw data. Most assesment-years have it as a numerical variable. The only exception so far is PASEC 1996, for which this variable is a string.

(c) Some assessment-years may not have the variable idclass.


Variables specific to a single assessment or year

Though the variable learner_weight exist in all assessments, other sample-related variables vary across assessments.

varnamevaluevarlabelvartypenote
yearkeyYear of assessmentDatePASEC, EGRA only (when multi-year bundles)
urban_o*traitOriginal variable of urbanCategoricalPIRLS, TIMSS, SACMEQ only (whenever available)
learner_weight_subject*sampleTotal learner weight for specific subjectNumericalLLECE only
strata*sampleStrataNumericalLLECE, PASEC only
jkzonesampleJackknife zoneNumericalPIRLS, TIMSS, PASEC 2014 only
jkrepsampleJackknife replicate codeNumericalPIRLS, TIMSS, PASEC 2014 only
weight_replicate*sampleReplicate weight #NumericalPASEC 2014 only

Variables specific to a single assessment or year

Though the variable learner_weight exist in all assessments, other sample-related variables vary across assessments.

varnamevaluevarlabelvartypenote
yearkeyYear of assessmentDatePASEC, EGRA only (when multi-year bundles)
urban_o*traitOriginal variable of urbanCategoricalPIRLS, TIMSS, SACMEQ only (whenever available)
learner_weight_subject*sampleTotal learner weight for specific subjectNumericalLLECE only
strata*sampleStrataNumericalLLECE, PASEC only
jkzonesampleJackknife zoneNumericalPIRLS, TIMSS, PASEC 2014 only
jkrepsampleJackknife replicate codeNumericalPIRLS, TIMSS, PASEC 2014 only
weight_replicate*sampleReplicate weight #NumericalPASEC 2014 only