PROC LIFEREG generates incorrect results under the following conditions:

• You specify the COEFFPRIOR=NORMAL(INPUT=SAS-data-set) option in the BAYES statement.
• The INPUT= data set specifies independent normal prior variances in observations where _TYPE_='VAR'.

In this scenario, incorrect results are produced in several places, including the posterior summaries, the posterior sample, and in the OUTPOST= data set.

To circumvent this issue, create observations in the INPUT= data set where _TYPE_='COV' and the off-diagonal elements are zero.

For example, the following INPUT= data set contains the means and variances of the normal prior:

In the code above, you can circumvent incorrect results by creating the INPUT= data set using _TYPE_='COV' instead of _TYPE_='VAR', as shown below:

Click the Hot Fix tab in this note to access the hot fix for this issue.

Operating System and Release Information

A fix for this issue for SAS/STAT 15.2 is available at: