Answers for "Non linear Regression with constraint SPSS (large standard deviations in results)"
https://developer.ibm.com/answers/questions/222479/non-linear-regression-with-constraint-spss-large-s.html
The latest answers for the question "Non linear Regression with constraint SPSS (large standard deviations in results)"Answer by maja7
https://developer.ibm.com/answers/answers/222481/view.html
<p>
Thank you for your reply! There is indeed a problem with multicollinearity when I look at the correlations. How could this influence the standard error? It is weird that in the paper I have read they use the same variables and they don't have so large standard error. The constraint in my model is that the sum of the estimated coefficients of 18 variables has to be equal to zero. How could I do this?
</p>Fri, 31 Oct 2014 02:41:14 GMTmaja7Answer by JonPeck
https://developer.ibm.com/answers/answers/222480/view.html
<p>
The large standard errors may be the result of a lot of collinearity among the regressors.
</p>
<p>
As for using linear regression, it is sometimes possible to impose a constraint by creating transformed variables that leave the problem as an unconstrained linear regression. For example, if two coefficients should be constrained to be equal, you can just enter their sum as a regressor. See also the TEST method of the METHOD subcommand or REGRESSION and the GLM procedure, which provides additional ways to do tests.
</p>Wed, 29 Oct 2014 21:03:40 GMTJonPeck