If you use the FLATDATA statement to transform variables and under the rare circumstance have more than 1000 time period observations for some cross sections, then the transformed values in the OUT= data set may be incorrectly duplicated multiple times starting from the 1000th time period observation forward. The procedure may or may not run out of memory in this case. If the procedure does not run out of memory, then the procedure will produce estimation results that are incorrect using the incorrectly transformed data.

If you have less than 1000 time period observations for all cross sections and if the time period index goes beyond 1000 for some or all the time periods, for example, x_1980, x_1981, etc., then the transformed values in the OUT= data set will be incorrectly transformed into missing values for those observations with time period index greater than 1000.

If the number of time series observations is less than 1000 for all cross sections, then you may circumvent the problem by changing the time period index to start from 1 so that all time period indices are less than 1000. If the number of time series observations is greater than or equal to 1000 for some cross sections, then you will need to transform the variables yourself prior to the PROC PANEL step to avoid using the FLATDATA statement.

Operating System and Release Information