Using weights with -movestay- for endogenous switching regression model

[Louis Hodge]

Dear all,

I am attempting to run estimation using the user written program -movestay- (ssc install movestay).

Using the following syntax the program executes successfully:

movestay y = x1 x2 x3 x4 , select(seldep = x1 x2 x3 x4 z1 z2 z3)
I would additionally like to include weights, according to the help file this is allowed as follows:

movestay y = x1 x2 x3 x4 [pw=weight], select(seldep = x1 x2 x3 x4 z1 z2 z3)
However, this returns a r(198) error. Has anyone else come across this problem or know why this would be?

Louis

[Marco Mello]

I have the same exact issue Louis.. Have you managed to find the solution to this problem associated with movestay?

[FernandoRios]

It seems this is a bug problem with the original movestay program.
I think the authors update can be installed using the following
"net install st0071_2.pkg"
That should fixed the weight problem.
HTH

[Marco Mello]

Hi FernandoRios, thank you very much for your information. I really hope that this update fixes the bug! Can you please explain to me how to install this? When I type net install st0071_2.pkg I get that the fille is not found..

file http://fmwww.bc.edu/RePEc/bocode/m/st0071_2.pkg not found
could not load st0071_2.pkg from http://fmwww.bc.edu/RePEc/bocode/m
r(601)

[FernandoRios]

My apologies,
The full command should be
net install st0071_2.pkg, replace from("http://www.stata-journal.com/software/sj5-3/")
if that doesnt work, use the following "findit st0071_2", and click on the first link it gives you to install the package
Best

[Marco Mello]

Thank you so much Fernando, it is true this modification fixes the bug and allows you to solve this weight issue.

Unfortunately, I believe it introduces another feature which was not present in the previous version of movestay. Now it does not allow me to have a control variable in the outcome equation which is not present in the selection equation. I can have controls in the selection which are not in the outcome (exclusion restrictions to improve identification) but not the opposite...

I have some variables affecting only the outcome and not the first stage estimation of the selection equation indeed..

[FernandoRios]

Hi Marco
My next suggestion is to estimate the model using the user written command oheckman. While technically it uses an order probit for the selection modeling, when you have only 2 options, the results are the same as a probit.
Fernando

[Antonio MartinsNeto]

Hi all,

I am having the same problem even though I installed the latest version of movestay. Have you guys managed to solve the problems with using weights?