SGP Data Analysis

Damian W Betebenner & Adam R Van Iwaarden

February 20th, 201i9


SGP analyses can be conducted in many ways with several variations. In general, one can always conduct the analyses using the lower level functions studentGrowthPercentiles and studentGrowthProjections. However, it is often easier to use higher level functions (wrapper functions) for the lower level functions, particularly for operational analyses. Both approaches to running SGP analyses are discussed below.

SGP analysis using lower level functions

SGP analysis using higher level functions

SGP Analysis in 6 Steps

  1. prepareSGP
  2. analyzeSGP
  3. combineSGP
  4. summarizeSGP
  5. visualizeSGP
  6. outputSGP

SGP Analysis in 1 Step

Because these steps are almost always conducted simultaneously in operational SGP analyses, the SGP Package has wrapper functions abcSGP and updateSGP that “wrap” the above 6 steps into a single function call, simplifying the source code associated with the analysis.

Running this script will run the six steps mentioned earlier by performing the following:

  1. prepareSGP Take exemplar LONG data, sgpData_LONG, and INSTRUCTOR-STUDENT lookup files, sgpData_INSTRUCTOR_NUMBER and create the SGP object Demonstration_SGP that includes them.
  2. analyzeSGP Performs SGP analyses for all years and content areas including: student growth percentiles, baseline student growth percentiles, student growth projections, baseline student growth projections, lagged student growth projections, lagged baseline student growth projections.
  3. combineSGP Merges results derived from analyzeSGP back into the master longitudinal record, Demonstration_SGP@Data, and calculates/creates scale scores associated with SGP targets.
  4. summarizeSGP Calculates summary tables and puts them results into the Demonstration_SGP@Summary slot.
  5. visualizeSGP Creates state level summary bubble plots, growth achievement plots (wave charts), and demonstration student growth plots.
  6. outputSGP Exports data in multiple formats:

The parallel.config argument is used to take advantage of computing resources with multiple processors/cores. In general, parallel computation works better on Linux/OSX than on Windows, though the package is tested on all three operating systems. Removing the parallel.config argument will result in the analyses being run sequentially — same results, just takes longer.

Contributions & Requests

If you have a contribution or topic request for this vignette, don’t hesitate to write or set up an issue on GitHub.