Dear all,
I am running a fractional logit model, using fracglm, where Vote (the percentages of votes in favor) is my dependent variable.
My independent variable of interest (X1) is a fraction, representing a group of people with voting power (1 vote per person). Hence, when X1 increases by 1, Vote should increase by a maximum of 1. The coefficient is relative to another group (not included in the regression), and both groups add up to the total voting power. Hence, the coefficient of X1 should be between -1 and 1.
However, X1 interacts with two additional variables:
Vote = a + [b1]X1 + [b2]X1 * X2 + [b3]X1 * X3
The total effect of X1 on Vote should be between -1 and 1 (-1<= b1 + b2 + b3 <=1).
How can I test such a model in Stata? I am aware of this FAQ but I am not sure whether it can be applied to my model.
Related Posts with Constraints in a nonlinear regression with interaction terms
Linear Regression based on Groups- monthly panel data; id average income, monthly average
- count models with random parameters (when the data is not panel data)
- Coefplot merging coefficients of multiple regressions by category
- The e(sample) function evaluates to 0 when it is not defined (and not to missing as I expected). Where is this documented?
Subscribe to:
Post Comments (Atom)
0 Response to Constraints in a nonlinear regression with interaction terms
Post a Comment