## Distribution for Proc Mixed Predictions

Occasional Learner
Posts: 1

# Distribution for Proc Mixed Predictions

I ran a model using the code below.  In the OUTP results I see a variable labeled StdErrPred, which I interpret as the standard error around the mu value (sume of random and fixed effects ) for a given observation.  I also have Covariance Parameters which create what I think of as the sigma squared result for the Normal distribution.  I think the two are related, but are not the same (the square root of the parameter estimate is not the same as the StdErrPred).  I cannot find a formula in the documentation that clearly describes what the StdErrrPred is.  The data set is a transformed lognormal data set; so the expected value of the observation in real dollars has to be created using the Mu and Sigma-Squared values; so, knowing with certainty what the correct Sigma Squared value is matters.  My interpration of the StdErrPred is that it reflects uncertainty around the estimate of the Mu and is not the same as the standard deviation of the variance of the probability distribution at a given value. Is that correct?

PROC MIXED DATA = WORK.SORTTempTableSorted

/* Start of custom user code */
COVTEST
/* End of custom user code */
PLOTS(ONLY)=ALL
METHOD=REML
;
CLASS cal_yr acc_yr_used acc_yr_roll Variance_Group_G Cal_Yr_Grp C_Dev_Time_20 C_Dev_Time_1 C_Dev_Time_2 C_Dev_Time_3 C_Dev_Time_4 C_Dev_Time_5 acc_yr_grp acc_yr_grp_2
;
BY cross_var;
MODEL ln_gross_pp_c= cal_yr_time Dev_Time_Cnt C_Dev_Time_10 C_Dev_Time_10*C_Dev_Time_10 C_Dev_Time_2 acc_yr_grp_2
/

/* Start of custom user code */
INFLUENCE
/* End of custom user code */
HTYPE=3
SOLUTION
CL
ALPHA=0.05
DDFM=KENWARDROGER
INTERCEPT
E3
OUTPM=WORK.MEAN_PP_TRAIN_GROSS(LABEL="Predicted means data set for WORK.TRAIN_GROSS_NEW")
OUTP=WORK.PRED_PP_TRAIN_GROSS(LABEL="Predicted values data set for WORK.TRAIN_GROSS_NEW")
;
RANDOM acc_yr_roll INT / CL
ALPHA=0.05 TYPE=VC;
REPEATED / SUBJECT=acc_yr_roll TYPE=VC GROUP=Variance_Group_G;
TITLE;
TITLE1 "Mixed Models Predictions";

Discussion stats
• 0 replies
• 193 views
• 0 likes
• 1 in conversation