https://communities.sas.com/t5/Statistical-Procedures/Beta-values-change-with-different-reference-groups-Glm/m-p/696384#M33611 Because the estimates are in reference to a standard, your reference group, and the estimate for the reference level itself is incorporated into the intercept.<BR /><BR /> If you change that reference point (the reference group) then the parameters will change to reflect the actual difference and the intercept will as well. This is correct and the expected behaviour. Your predicted values should not change.<BR /><BR /> Wed, 04 Nov 2020 00:25:24 GMT Reeza 2020-11-04T00:25:24Z https://communities.sas.com/t5/Statistical-Procedures/Beta-values-change-with-different-reference-groups-Glm/m-p/696367#M33610 <P>I was conducting GLM regression with log link and gamma distribution for modeling costs.&nbsp; I see that beta value for intercept changes with change in the reference groups for the covariates.&nbsp; Why does this happen?&nbsp; Can someone explain this?</P> Tue, 03 Nov 2020 23:18:24 GMT https://communities.sas.com/t5/Statistical-Procedures/Beta-values-change-with-different-reference-groups-Glm/m-p/696367#M33610 am12scorp 2020-11-03T23:18:24Z https://communities.sas.com/t5/Statistical-Procedures/Beta-values-change-with-different-reference-groups-Glm/m-p/696384#M33611 Because the estimates are in reference to a standard, your reference group, and the estimate for the reference level itself is incorporated into the intercept.<BR /><BR /> If you change that reference point (the reference group) then the parameters will change to reflect the actual difference and the intercept will as well. This is correct and the expected behaviour. Your predicted values should not change.<BR /><BR /> Wed, 04 Nov 2020 00:25:24 GMT https://communities.sas.com/t5/Statistical-Procedures/Beta-values-change-with-different-reference-groups-Glm/m-p/696384#M33611 Reeza 2020-11-04T00:25:24Z