3SLS and Limited Dependent Variable

[Ritika Khurana]

Hello,

I'm trying to estimate a 3-equation model in which my first stage dependent variable is a count, second stage variable is continuous and third (final stage) is limited dependent variable (lies between 0 and 100)

Is there a command or procedure that could address above mentioned type simultaneous system. For a basic (3SLS) simultaneous panel, I was using -reg3- with country fixed effects. But i found my estimates were incorrect as my third stage dependent variable ranges between 0 and 100.
I am following two papers on Nelson and Olson (1978) and Blundell and Smith (1989) method of Two-Stage LDV.
On searching I found out -gsem- and -cmp- . But even, these don't seem to address my problem.

Could you please help me with this?

Thank you

Ritika

[Phil Bromiley]

You didn't get a quick answer. You'll increase your likelihood of a useful answer by following the FAQ on asking questions - Stata code in code delimiters, readable Stata output, and sample data using dataex..

GSEM and cmp are the likely procedures. You don't say why they don't work for you.

[Ritika Khurana]

Thank you Phil for getting back to me. I will be elaborate more.

Since, I couldn't find my solution anywhere. I tried running two different model one 3sls and the other 2sls. And my equations go like this.

For 3sls,

occ_fl = avgrain avgtemp i.cty i.year
ptaf_fl = occ_fl i.cty i.year
frleg = occ_fl ptaf_fl lcgdp aid i.cty i.year

occ_fl is a count variable
ptaf_fl is a continuous variable
frleg is a fractional variable ranging between 0 and 1.

I found out cmp command, but that doesn;t take care of -zinb- i.e. my equation as there are many zeros.
I couldn't find a solution to this. I changed my equations and story and thought of trying 2sls instead

and this is what I ran.

cmp (ptaf = occ_fl occ_st lcgdp aid ) (frleg = ptaf occ_fl occ_st lcgdp aid i.cty i.year), ind($cmp_cont $cmp_frac)

I tried this equation with and without fixed effects in first stage of 2sls to see if it works.

where, ptaf = total no. of people affected by all disasters normalized by population.
Occ_fl = occurrence of flood, occ_st = occurrence of storms, lcgdp = log of per capita GDP, frleg = score of economic freedom (0 to 1), icty = country fixed effect, i.year = year fixed effect.

When I run this model, it doesn't seem to converge and continuously says (not concave) and it runs like this for 40 mins.

Now, my questions are: 1). If I want to use 3sls when which command will address it?
2). Is there a way to get solution for 2sls where the model converges?

Thank you.

Ritika