Using PARSCALE via Stata and Dans spreadsheet

1
Using PARSCALE via Stata and Dans spreadsheet

• Laura Gibbons, PhD

2
What does PARSCALE do?

• Estimates the unidimensional latent trait from
  the test items.
• (Rich will show you how to assess dimensionality
  tonight.)

3
Useful output files SCO

• RUNNAME.SCO (where RUNNAME is the name you assign
  the job)
• Score files have
• an estimate of theta (the latent trait) for each
  person (line of data).
• the standard error of that estimate. Remember
  that not all subjects are estimated with equal
  precision.

4
Useful output files PAR

• RUNNAME.PAR
• Parameter files give the difficulty and
  discrimination parameters for each item.

5
Not so useful output files

• RUNNAME.FIT (PARSCALE fit statistics are
  generally too conservative)
• Internal to PARSCALE, we never use
• code_runname.PH0
• code_runname.PH1
• code_runname.PH2
• code_runname.PH3
• pscal

6
Why Stata?

• You can use PARSCALE alone, but the input files
  are cumbersome.
• Our user interfaces make it much simpler
• runparscale (Calculates theta and merges thetas
  and SE back into your data set. Among other
  things)
• prepar (Just prepares the input file and data set
  for PARSCALE.)

7
Using Stata

• Now well switch to Stata. Ive saved many of
  the commands in .do files for you.
• Best to open Stata from the Start menu.
• (explain the 4 windows, including saving a log
  file)

8
Stata Do files

• Open a .do file from the envelope thats right
  below the USER pull down.
• Open Preliminaries for PARSCALE workshop.do

9
runparscale, prepar

• Open making thetas in PARSCALE.do

10
PARSCALE after prepar

• File open
• Run run all
• Run plot
• PARSCALE numbers your items in order
• Show ICC and Info and Total Info
• To see another type of plot, click on main menu

11
You try it

• Repeat all this for the memory scale
• llcorrt1_r, llcorrt3_r, lrcorr, lrcuecor
• If time, try the executive functioning scale
• digitbwd, visbwd, vftotalf_r, vftotala_r2,
  vftotals_r, drsip_r

12
Dans amazing spreadsheet

• To see the effect on the information curve/sem
  when you include or exclude items

13
Inputting the parameter file

• From within Excel, File-gtOpen (all files)
• Open GLOBAL.PAR
• Click on
• Delimited
• Starting at row 6
• Next
• Space
• Next
• Finish

14
IRT Scale Summary.xls

• Enable macros
• Copy the Excel cells from GLOBAL.PAR into cell A3
  in the Parameters sheet.
• Go to Row 184 and click
• In the TIC-ICC sheet, put a 1 above all items you
  want to include. Change to 0 to remove an item.
• Note that you will have n-1 columns if an item
  has n response categories

15
And viola!

• Go to GM to GU 14 and below to see the curves!
• Add or subtract items as desired.