topic Re: interpolating 100+ variables in SAS Forecasting and Econometrics

interpolating 100+ variables

putteringpluie — Tue, 13 Feb 2018 18:44:05 GMT

I have census data from 2001 and 2011. I want to interpolate all the variables for the in-between years by simple linear function. I don't want to use proc expand because of its constraint that slope will be zero in the beginning and end values. Any suggestions?

  proc expand data=have out=want;
      by state;
      id year;
   run;

Re: interpolating 100+ variables

Reeza — Tue, 13 Feb 2018 18:45:08 GMT

PROC REG or a linear regression type models? That's still going to be a hell of a lot of variables and the assumption of constant growth isn't correct either.

Re: interpolating 100+ variables

putteringpluie — Tue, 13 Feb 2018 19:06:52 GMT

Thank you for such a quick reply. I was wrong about the assumptions in proc expand. The slope need not be zero.

Re: interpolating 100+ variables

Reeza — Tue, 13 Feb 2018 19:24:37 GMT

PROC TIMESERIES will also fill in missing time with several options.

Re: interpolating 100+ variables

ballardw — Tue, 13 Feb 2018 21:02:17 GMT

@putteringpluie wrote:

I have census data from 2001 and 2011. I want to interpolate all the variables for the in-between years by simple linear function. I don't want to use proc expand because of its constraint that slope will be zero in the beginning and end values. Any suggestions?

If this is US Census bureau data I would recommend using the annual estimates that are available. I doubt if anything the Census works with actually has an 11-year straight line trend except the year variable.