eat Packages

eatAnalysis

Various useful help functions, such as saving Excel files, saving analysis results from lm4 and simulating IRT responses.

Benjamin Becker, Sebastian Weirich, Karoline Sachse

eatATA

Automated block occupation/automated test booklet creation.

Benjamin Becker

eatCodebook

Automated creation of scale manuals.

Benjamin Becker

eatFDZ

Automated anonymisation of data sets, matching of pdf documents (e.g. scale manuals) and data sets.

Benjamin Becker

eatGADS

Allows import and data preparation of SPSS data sets in R. Generates the General Analysis Data Set (GADS) for IQB Bildungstrend studies as a SQLite3 database. Parts of the data set can then be loaded into R using the package. It also allows the export of SPSS files to and from R.

Benjamin Becker

eatModel

Serves as an interface for the ConQuest software. The required control files (script, labels, data set in ‘fixed width’ format) are automatically generated and ConQuest is called via the command line. The resulting files (showfile, WLEs, PVs, etc.) can be imported back into R and edited further. Newer versions of ‘eatModel’ also allow the integration of the R package ‘tam’ and parallelisation.

Sebastian Weirich

eatPlot

Creating plots from the eatRep output (main use: Bildungstrend).

Nicklas Hafiz

eatPrep

Preparation of test item data for IRT scaling using metadata provided by a predefined list structure as well as several tools and checks. Raw data can be read in and will be subjected to plausibility checks. Multiple data sets can be merged in a single step, incorporating detailed diagnostics for non-identical values among identical cases and variables. Data will be recoded according to the provided meta data, aggregated and scored. Handling different types of missing data is customizable according to user specifications. Further, missing patterns can be checked, various rater agreement measures and category discriminations can be calculated. The input metadata can be generated by the IQB-internal tool ‘AnalyseInput’, which extracts item and test design information from the IQB database and formats it into a standardized xlsx file. This file can then be read by functions specifically designed for this task. Alternatively, the metadata can be provided in R-specific data formats.

Karoline Sachse

eatRecode

Create and apply recoding databases.

Benjamin Becker, Nicklas Hafiz

eatRep

Calculates means, standard deviations, variances, frequency tables, percentiles and linear (logistic) regressions, as well as trends for all these analyses in clustered multilevel structures with imputed data. The package implements part of the functionality of the WesVar computer software in R and is mainly relevant for the IQB Bildungstrend studies.

Sebastian Weirich, Benjamin Becker

eatTools

Various help functions that are also required by the packages ‘eatPrep’, ‘eatModel’, ‘eatGADS’ and ‘eatRep’, among others.

Sebastian Weirich, Benjamin Becker, Karoline Sachse
No matching items

Installing R packages

Detailed instructions for installing the R packages can be found here: installing R packages

Compatibility

Most packages of the “eat” family are not executable in isolation, which means that a package is usually dependent on another package in order to be executable. This means:

  • “eatTools” is a package with help functions required by other packages. “eatTools” does not depend on other “eat” packages.
  • “eatPrep” requires “eatTools”.
  • “eatRep” needs “eatTools” and “eatGADS”.
  • “eatModel” requires “eatTools” and “eatRep” (and thus also “eatGADS”). The package optionally accesses the computer programme Conquest or the R package “TAM” for parameter estimation. For Conquest, the command line executable is sufficient, e.g. “console_Feb2007.exe”; “TAM” is installed when the package is loaded. “TAM” and Conquest are based on the same statistical measurement model, and overlap considerably in their range of functions.
  • “eatGADS” requires “eatTools” and “eatDB”.

The interdependence also extends to different versions of the packages. For example, old versions of “eatRep” are not always compatible with new versions of “eatTools” (and vice versa). The current package versions (see table above) should be compatible with each other. Usually (but not always) version conflicts are indicated by an error message. It is recommended to always have the latest package versions installed.

Since individual functions have also changed substantially in the course of the package development, it may no longer be possible to replicate past analyses with new package versions and the identical script (e.g. from the Ländervergleich 2011). In this case, the old script would either have to be adapted or the package versions used at that time would have to be restored.