In Table 20.1

Explanation

TABLE OF MEASURES ON TEST OF 13 ITEM

Raw score-to-Measure Table for all raw scores on the complete set of all active non-extreme calibrated items. This can also be output with SCOREFILE=.

Table 20.1 shows the score-to-measure table for persons with complete data. For persons with missing data, this table does not apply.

Score-to-measure tables for subtests or persons with missing data can be produced by using ISELECT= or IDELETE= from the Specification menu before requesting Table 20.

If there are subsets, then the reported Table 20 is one of an infinite number of possible Table 20s.

If you want the score file for person measures including the extreme (zero, perfect) items, then

1. Run a standard analysis.

2. Output: IFILE=if.txt SFILE=sf.txt
Eliminate unwanted items from IFILE= and remove any ;

3. Run the analysis again, with Extra specifications: IAFILE=if.txt SAFILE=sf.txt

4. The person measures will have altered somewhat to adjust for the imputed difficulties of the extreme items.

5. Output Table 20 and SCOREFILE=

SCORE

Raw score on the complete set of non-extreme items.  If IWEIGHT= produces decimal scores, then scores with one decimal place are shown. Some SCOREs may not be observable. Some observable scores may not be shown, but may be better approximated from TCCFILE=

MEASURE

Estimated measure for the SCORE raw score on all the items, adjusted by USCALE= and UIMEAN=, if active. The MEASURE may differ slightly from Table 17. Tighten convergence to make them agree exactly.

The SCOREFILE= and Table 20 person ability estimates are estimated on the basis that the current item difficulty estimates are the "true" estimates. These are the person estimates if you anchored (fixed) the items at their reported estimates. The convergence criterion used are (LCONV=)*.01 and (RCONV=)*.01 - these are considerably tighter than for the main analysis. So Table 20 is a more precise estimate of the person measures based on the final set of item difficulties.

PFILE=, Table 17 and the other Person Measure Tables show the person abilities that are the maximum likelihood estimates at the current stage of estimation. To make these two sets of estimates coincide, please tighten the convergence criteria in your Winsteps control file:

CONVERGE=L

LCONV=*.001 ; or tighter

If STBIAS=YES is used, then score table measures are dependent on the data array, even if items are anchored.

-6.34E

"E" (Extreme, Extrapolated) is a warning that the accompanying value is not estimated in the usual way, but is an approximation based on arbitrary decisions. For extreme scores, the Rasch estimated would be infinite. So the arbitrary decision has been made to make the extreme score more central by the EXTREMESCORE= amount (or its default value), and to report the measure corresponding to that score. The Rasch standard error of an extreme score is also infinite, so the reported standard error of an "E" value is the standard error of the finite reported measure.

S.E.

The model-based standard error of the MEASURE, not adjusted for misfit.

CURRENT VALUES, UMEAN=.0000 USCALE=1.0000

This shows the current user-scaling. Changing the user-scaling in the Specification menu dialog box or the Help menu Scaling calculator changes Table 20 immediately. Use this to experiment with different values of USCALE= and UIMEAN=.

TO SET MEASURE RANGE AS 0-100, UMEAN=46.5329 USCALE=7.3389

Suggested UIMEAN= and USCALE= values for transformation of the MEASUREs into user-friendly values in the range 0-100 resembling percentages.

TO SET MEASURE RANGE TO MATCH RAW SCORE RANGE, UMEAN=49.2956 USCALE=5.7243

Suggested UIMEAN= and USCALE= values for transformation of the MEASUREs into user-friendly values into the range or raw scores so that the  MEASUREs resemble raw scores.

Predicting Score from Measure: Score = Measure * 8.2680 + 38.9511

A linear approximation of the score-to-measure ogive useful for predicting raw scores on this set of items from measures.

Predicting Measure from Score: Measure = Score * .1158 + -4.5099

A linear approximation of the score-to-measure ogive useful for predicting measures from raw scores on this set of items.

Statistically different levels of performance = 3.8 Wright's Sample-independent Person (Test) Reliability = 0.94

Across the raw-score range in Table 20.1, statistically significant different raw scores are 1.96 * (M1 - M2) / √(SE12 + SE22) apart where M1 and M2 are the two measures. SE1 and SE2 are their two standard errors. See www.rasch.org/rmt/rmt144k.htm. The number of different levels resembles a separation index. The matching "Test" Reliability is shown. These are equivalent to the Separation and Reliability of a person sample with a uniform distribution having the range of the MEASUREs in Table 20.