In SPSS, how to obtain a single prediction of the dependent variable for each fixed factor, rather than for each random factor combination? I have a number of thing-types X and for each type want to obtain a single estimate of its true value Y.
When I have measured the same thing-type several times in the same session and several times each in sessions on different days, then if I understand correctly the thing-type is a fixed factor and the absolute day is a random factor (in that I could hypothetically keep doing more measurements for different days, each day influencing the results by introducing a different bias).
When using SPSS to try to obtain estimates, I was elated when seeing the options of saving predicted values and confidence intervals to the dataset.
I put in thing-types as Subjects, and in Random Effects added the different random factors as blocks.
However, when I looked at the output spreadsheet, I saw that measurements with all the same factors had the same predicted value, but measurements with the same fixed factor and different random factors had different predicted values.
At the time, I was attempting to use the Mixed Model Generalized Linear option for Analyze.
Searching near-blindly for the way to get what I imagine without knowing the correct terminology has not been going well.
Is there a way to do what I am imagining?  To put in measurements of several things under several conditions, and to get out one estimate for each thing?
Thank you for your time.
 A: The MIXED procedure, which fits linear mixed models, will save both predicted values, which include fixed and random effects, and also fixed predicted values, which include only the fixed effects. The GENLINMIXED procedure only saves predicted values, but it is possible to save the model out and then use the Scoring Wizard to score your data, and this produces the fixed predicted values.
When you run the generalized linear mixed model, go to the Model Options tab's Save Fields screen, where you can save predicted values. In the bottom of the screen you can choose to export the model in .xml format. Choose a file location and run the procedure. 
Then, in the menus choose Utilities>Scoring Wizard, navigate to the model file you just saved, and select it. If you've saved the predicted values in GENLINMIXED, you'll have to change the name of the variable into which to save the fixed predicted values, as the default is the same "PredictedValue" string. Clicking Done should produce fixed predicted values from the generalized linear mixed model.
