COMMENT	spss code for Example 3 .
COMMENT	 This is the final model .
COMMENT	Initial is the Car versus Salesperson prime.
COMMENT	NFCC is the mean of the need for certainty and need for closure scores.
COMMENT	ctrNFCC is NFCC transformed to a z-score variable. 
MODEL PROGRAM B01 = -0.3 G01 = -1.0 Q1 = 0.25 Q11 = -0.6 Q2 = -0.7 Q12 = 0.5 Q13 = -0.5  .
COMPUTE	#DV = prob .
COMMENT	Composition submodel .
COMPUTE	#C1 = EXP(Q1 + Q11*Initial + Q12*ctrNFCC + Q13*Initial*ctrNFCC)/(1+EXP(Q1 + Q11*Initial + Q12*ctrNFCC + Q13*Initial*ctrNFCC)+EXP(Q2 - Q11*Initial - Q12*ctrNFCC - Q13*Initial*ctrNFCC)) .
COMPUTE	#C2 = EXP(Q2 - Q11*Initial - Q12*ctrNFCC - Q13*Initial*ctrNFCC)/(1+EXP(Q1 + Q11*Initial + Q12*ctrNFCC + Q13*Initial*ctrNFCC)+EXP(Q2 - Q11*Initial - Q12*ctrNFCC - Q13*Initial*ctrNFCC)) .
COMMENT	Location submodel .
COMPUTE	#M1 = EXP(B01)/(1+EXP(B01)) .
COMPUTE	#MU = (1-#C1-#C2)*#M1 + #C1/2 + #C2/4 .
COMPUTE	#M2 = 1/2 .
COMPUTE	#M3 = 1/4 .
COMMENT	Dispersion submodel .
COMPUTE PHI1 = EXP(-G01) .
COMPUTE	#D = .01 .
COMPUTE PRED_ = #MU .
COMPUTE RESID_ = #DV  - PRED_ .
COMPUTE LL = LN(MAX(0.001, (1-#C1-#C2)*PDF.BETA(#DV, #M1*PHI1, MAX(0.001, PHI1 - #M1*PHI1))+ #C1*PDF.UNIFORM(#DV, #M2-#D, #M2+#D) + #C2*PDF.UNIFORM(#DV, #M3-#D, #M3+#D))) .
COMPUTE LOSS_  = -LL .
COMMENT	To get bootstrap estimates, put /BOOTSTRAP n statement before /CRITERIA statement.
COMMENT	To save predicted values, residuals and gradient put /SAVE PRED RESID DERIVATIVES before /CRITERIA statement.
CNLR prob
  /OUTFILE='C:\SPSSFNLR.TMP'
  /PRED PRED_
  /LOSS LOSS_
  /CRITERIA STEPLIMIT 2 ISTEP 1E+20 .