While Winsteps is running, information about the analysis is displayed on the screen. The iterative estimation process is by logistic curve-fitting. Here is an example based on the "Liking for Science" data, example0.txt

 

 

The Winsteps analysis window can display:

 

WINSTEPS  Version: 5.7.4

program running - shows version number

 

Reading Control Variables .. Processing your control variables

Input in process ......   Reading in your data: each . is 1,000 persons

 

1211102012222021122021020  M Rossner, Marc Daniel

^I                      ^N ^P

^I where item response string starts, ITEM1=

^N last item response, NI=

^P where person label starts, NAME1=

 

 75 person records input  Total person records found

 

To stop iterations: press Ctrl+S

To cancel the analysis: press Ctrl+F

 

If Winsteps detects that the iterative process is not converging (the maximum residual and maximum change are not becoming smaller) then this message appears:

"Estimation may have diverged. Press Ctrl-F to stop iterations."

Reasons for this include:

1.Incorrect values in the data. Look at CODES= in your Winsteps control file, and Table 3.2.

2.The data form disconnected subsets.

 

CONVERGENCE TABLE: See Table 0.2

 

PROBING DATA CONNECTION: See SUBSETS=

 

PROX and JMLE: See Estimation methods

Check that values diminish, and become near 0. See convergence considerations

 

When CMLE=Yes, CMLE estimation is done:

 

 

JMLE estimation continues .....

 

 

CALCULATING FIT STATISTICS

During this iteration through the data, the person and item fit statistics are computed. The exact formulas are in the book Rating Scale Analysis, p. 100 - also at Fit Equations.

 

Summary statistics are extracted from Table 3.1 with REALSE=Yes and including extreme scores.