If a residual/response output filename is specified by Residual file=filename, a file of responses and residuals from the main analysis is produced. This file is designed for input into other programs with one line per measurable observation. It can be used for calculating other fit statistics and producing specialized diagnostic reports.

 

Heading lines= control the output of the heading line. CSV= allows tab-delimited and other formats. QM quotation marks, controls whether labels are within quotation marks.

 

filenames ending .xls and .xlsx are written as Excel workbooks

filenames ending .sav are written as SPSS save files

filenames ending .rda and .rdata are written as R Statistics data files

filenames ending with any other suffix or no suffix are written as DOS text files.

 

This file can be produced from the Output Files menu by clicking on Residual/Response Output file. This has additional options.

 

Here is an example of the format with 4 decimal places in the "Select fields" dialog box. The precise format depends on the number of facets in your data:

 

       Obs       Stp       Exp       Res       Var     StRes        Wt     LProb   Measure     Displ    Status     MPCat        E1        E2        M1        M2 Children  Tapping_it

         1         1       .97       .03       .02       .16      1.00      -.03      3.66      1.03         2         1         1         1     -2.98     -6.64 Boy       1-4      

         1         1       .97       .03       .02       .16      1.00      -.03      3.66      1.03         2         1         1         2     -2.98     -6.64 Boy       2-3      

 

The columns are:

 

 

For "Category implies Measure" (C->M) and "Measure implies Category" (M->C) statistics, for each observation in the Facets Residualfile=,

"expected score for this response" - round this to the nearest category number = expected average category

if "expected average category" = "observed response as renumbered into a count of ordered steps" then MC = 1, else, MC = 0.

Compute average of MC for each observed category across all the relevant data for C->M

Compute average of MC for each expected category across all the relevant data for M->C

 

Example: The "Obs" (observed) is the original data. The "Stp" (step) is the ordinal version of the original data. This version is used for analysis, and is the version on which the "Exp" (expected) and the "Res" (residual) are based. This version may be the same as the original data, or the original data may be transformed either due to explicit instructions by the analyst, or by default operation of Facets.

 

For instance, suppose that the original data are observations of these three values: 10, 20 and 30. Then, by default, Facets will analyze these observations as the "steps": 10, 11, 12. If the original data are intended to be 10,11,12,13,14,....,28,29,30. Then please specify this is in your Models= statement:

Models=

?,?,..., R30K   ; where "K" means "Keep" the original numeration.

 

Example 1: I need the the S.D. of the observed ratings for each rater.

Facets does not output this statistic. So, output the Residualfile= to Excel. Sort by Rater number. Then compute the S.D. for each rater separately or use the Excel SUBTOTAL function:

1.On the Data tab, in the Outline group, click Subtotal. The Subtotal dialog box is displayed.

2.In the At each change in box, click the nested subtotal column. ...

3.In the Use function box, click the summary function that you want to use to calculate the subtotals. ...

4.Clear the Replace current subtotals check box.

5.Click OK

 

Example2 : Where I can find the adjusted ratings after incorporating the differences in the leniency/severity measures on the raw ratings of the corresponding raters?

 

1. Estimate your data using Facets.

Output an Anchorfile= with data.

2. In the anchored file, change the anchored rater severities to 0 logits. Keep everything anchored.

3. Analyze the anchored file with Facets. Everything should be anchored, and there should be displacements.

4. Output the Residual File. The "expected score for this response" is the adjusted rating.

 

Akaike Informaton Criterion: AIC = 2k - 2ln(L)

where k = number of free parameters, and ln(L) =  log-likelihood of the data.

ln(L)  = the sum of the Lprob field in the Residual file.  From the Facets "Output Files" menu, output the Residual file to Excel, then sum the LProb column.

The number of free parameters is difficult to estimate. You could use the total number of elements as an approximation.

 

Bayesian Information Criterion: BIC = ln(n)k - 2ln(L)

where n is the number of observations = rows in the residual file