For each cross section, PROC TSCSREG requires at least two time series observations with nonmissing values for all model variables. In addition to this requirement, for each time point, PROC TSCSREG requires at least two cross sections with nonmissing values for all model variables. If either of these two conditions is not met, then the following error message might appear in the log:

   ERROR:  There is only one cross section or time series
   observation.  Computations will be terminated.

In investigating the specific cause of such an error, it might be helpful to first identify those observations for which the values of at least one of the model variables is missing. Create a variable _TEMP_ = Y + X1 + X2 + ... + X20, which is the sum of all the model variables. The value of _TEMP_ is missing whenever one or more of the model variables has missing values. Use PROC MEANS to identify how many such observations exist in the data set. For the FIXONE and FIXTWO methods, cross sections with only one time series observation provide no additional information to improve estimation of the regression coefficients. To run PROC TSCSREG successfully for these methods, first delete observations in the input data set for the offending time point or cross section. Or the TSCSDATA macro might be very helpful in removing such observations when one or both of the input data requirements are violated.

For the RANONE and RANTWO cases, you might be able to use such cross sections (that have only one time series observation) to improve the estimate of the regression parameters by using PROC MIXED, which does not impose such a requirement on the input data set. Or use the TSCSDATA macro to obtain a subset of the data for which TSCSREG runs successfully.

Operating System and Release Information