I'm trying to use PROC TSCSREG to do a regression analysis on some panel data; I'm wanting to do it as fixed error.
The code I've written is as follows (variable names changed to protect the innocent):
PROC SORT DATA= project;
BY csid tsid;
RUN;
PROC TSCSREG DATA= project;
ID csid tsid;
MODEL y= x1 x2 x3/ fixone;
RUN;
when I run this, SAS returns an error message:
"ERROR: There is only one cross section or time series observation. Computations will be terminated."
there is not, however, only one observation, there are many. If I switch the csid and tsid (putting time in for cross section), then the procedure runs just fine; but of course tells me I have the number of time observations in individuals, and the number of individuals in time observations.
I've checked, and there is NO missing data on the explanatory variables. However, there is a fair amount of missing data in the response variable.
additionally, a lot of the cross sectional id's are non-sequential; thus instead of id numbers such as "1, 2, 3, 4, 5, ..." it goes "1, 3, 7, 11, 45, ...". However, the first three id numbers are 1, 2, and 3; meaning if it is the non sequential id'ing, then it should still be getting more than "only one ts or cs observation".
That's all the relevant information I can think of on this; not sure what else to include that might be causing the problem. I'm totally stumped. Anyone got an idea why this is happening?
The code I've written is as follows (variable names changed to protect the innocent):
PROC SORT DATA= project;
BY csid tsid;
RUN;
PROC TSCSREG DATA= project;
ID csid tsid;
MODEL y= x1 x2 x3/ fixone;
RUN;
when I run this, SAS returns an error message:
"ERROR: There is only one cross section or time series observation. Computations will be terminated."
there is not, however, only one observation, there are many. If I switch the csid and tsid (putting time in for cross section), then the procedure runs just fine; but of course tells me I have the number of time observations in individuals, and the number of individuals in time observations.
I've checked, and there is NO missing data on the explanatory variables. However, there is a fair amount of missing data in the response variable.
additionally, a lot of the cross sectional id's are non-sequential; thus instead of id numbers such as "1, 2, 3, 4, 5, ..." it goes "1, 3, 7, 11, 45, ...". However, the first three id numbers are 1, 2, and 3; meaning if it is the non sequential id'ing, then it should still be getting more than "only one ts or cs observation".
That's all the relevant information I can think of on this; not sure what else to include that might be causing the problem. I'm totally stumped. Anyone got an idea why this is happening?