* Chapter 7 Model Diagnostics in SPSS .
* Fit Model 7.3, and save conditional predicted values and conditional residuals .
MIXED
gcf WITH time base_gcf cda age
/CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE)
PCONVERGE(0.000001, ABSOLUTE)
/FIXED = time base_gcf cda age | SSTYPE(3)
/METHOD = REML
/PRINT = G SOLUTION TESTCOV
/SAVE = PRED RESID
/RANDOM INTERCEPT time | SUBJECT(Patient) COVTYPE(UN)
/RANDOM INTERCEPT | SUBJECT(tooth* Patient) .
* Plot conditional residuals versus conditional predicted values (Figure 7-4) .
* NOTE: Standardized residuals are not currently available automatically in SPSS .
GRAPH
/SCATTERPLOT(BIVAR)=PRED_1 WITH RESID_1
/MISSING=LISTWISE .
* NOTE: A reference line at zero can be manually added to the above plot by double-clicking on the plot to edit it .
* Generate a Q-Q plot to assess the conditional residuals for normality (Figure 7-5) .
PPLOT
/VARIABLES=RESID_1
/NOLOG
/NOSTANDARDIZE
/TYPE=Q-Q
/FRACTION=BLOM
/TIES=MEAN
/DIST=NORMAL.
* NOTE: plots of the individual EBLUPs in a model with multiple random effects cannot currently be generated in SPSS,
* since the individual EBLUPs cannot be saved in the data set.