PROC SPATIALREG returns "This model cannot be fit", but why?

[fpb1]

I am trying to relate Medicare spending to various census variables nationwide. I am using zip code level data, so the number of spatial units is over 20,000. The error message is not informative enough to tell me if it's a memory issue or a matrix algebra issue or something else. It does work for a 10% subset of the data, which makes me lean toward a memory issue, but in my experience if it's a memory issue SAS returns a more explicit error message.

Here is the log:

model var2 = var1 / type=SAR;

spatialID zip;

quit;

ERROR: This model cannot be fit.

NOTE: PROCEDURE SPATIALREG used (Total process time):

real time 1:20.53

user cpu time 1:20.47

[ballardw]

When you did the test with 10% of the data was it a random sample from the entire database or perhaps a more geographically homogeneous (i.e. all the data from a few states)?

How many actual records did you use?

Your log is missing the actual PROC Statements and all the options that might be on that statement.

---

@fpb1 wrote:

I am trying to relate Medicare spending to various census variables nationwide. I am using zip code level data, so the number of spatial units is over 20,000. The error message is not informative enough to tell me if it's a memory issue or a matrix algebra issue or something else. It does work for a 10% subset of the data, which makes me lean toward a memory issue, but in my experience if it's a memory issue SAS returns a more explicit error message.

Here is the log:

model var2 = var1 / type=SAR;

spatialID zip;

quit;

ERROR: This model cannot be fit.

NOTE: PROCEDURE SPATIALREG used (Total process time):

real time 1:20.53

user cpu time 1:20.47

---