In both PROC TSCSREG and PROC PANEL if you have any cross sections with all missing values then the following error message will appear in the SAS Log:

• ERROR: Floating Point Zero Divide error.

To circumvent the floating point zero divide error delete those cross sections with all missing values.

In PROC PANEL if you have any cross sections with only 2 observations and you see the following:

1. FIXONE model
2. Lagged dependent regressor

Then the following error message will appear in the SAS Log:

• ERROR: Floating Point Zero Divide error.

Under these conditions there is only a single observation available to estimate the model for those cross sections with 2 observations. Rather than fail with this error, the procedure should produce a note which indicates that the first cross section has insufficient data.

To circumvent the problem, either use PROC TSCSREG rather than PROC PANEL to fit the one-way fixed effects model with the lagged dependent regressor or remove all cross-sections with only 2 observations and rerun the analysis in PROC PANEL.

Operating System and Release Information