Home/ Questions/Q 8288947
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-08T12:23:21+00:00

Possible Duplicate: (cross-post from stats.SE, but I was sent here by a mod instead)

  • 0

Possible Duplicate: (cross-post from stats.SE, but I was sent here by a mod instead)
How to get SSE for predictions using SAS?

enter image description here

I’m trying to get the sum of squared errors for my predictions in SAS, but I’m not sure I’m doing it correctly. I’m not sure I fully understand the output I’m getting with my code (specifically, stdp):

data tridata;
infile '\data.dat';
    input x1 x2 x3 y;
    proc sort data = tridata; by x3;

proc reg data = tridata;
    model y=x3;
    plot r. * x3;
    output out = tridata2 r = resid p = pred stdp = err;
run;
quit;

/* Send your errors to a file */
data _NULL_;
    file '\data-err.dat';
    set tridata2;
    put err;
    where y eq .;
run;
quit;

This gives me a file of errors for each estimate. I imported these into Excel, squared each one, and summed them up to give me a number. Is this the correct method? Based on the description of my project, I was under the impression that I should get an SSE_test for each predicted value. See below:

enter image description here

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    First of all , STDP is Standard Error of the mean predicted value[this is used in calculating 95% confidence/prediction intervals]. This is not the one you need to square and sum up to get the SSE.

    This is what you need to do:

    Where you had “r = resid” in the output statement in PROC REG is the measure of your error. It is simply Actual – Predicted. Since since sometimes your model may over-predict and some times under predict. This R, Residual, could be a positive value or a negative value.

    In order to make everything positive, we just square the residual values and get SQUARED ERRORS for each of the predictions you’ve made.

    Sum up all the SQAURED ERRORS and you get SUM OF SQURED ERRORS. This will be a single statistic(or a scalar).

    As far as I know, you cannot calculate your SSE unless you have the actual values to hand.

    You may refer to http://en.wikipedia.org/wiki/Errors_and_residuals_in_statistics specifically the last paragraph on that page.

    • 0