If you request a kernel-smoothed hazard plot and confidence limits with the PLOTS=HAZARD(CL) option in the PROC LIFETEST statement, the displayed confidence interval is incorrect. Similarly, if you save the resulting hazard plot data using the statement ODS OUTPUT HazardPlot=data-set, the UPPER and LOWER variable values are incorrect. The problem occurs for the METHOD=BRESLOW, FH, or KM options in PROC LIFETEST. The problem also occurs for any hazard-option KERNEL=BIWEIGHT, EPANECHNIKOV, or UNIFORM.

By default, METHOD=KM or its alias, PL, and when you specify PLOTS=HAZARD the default KERNEL= is Epanechnikov.

There is no workaround for this problem.