Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Absolute beginners guide to Multivariate probit analysis using Stata 12

From   "JVerkuilen (Gmail)" <>
Subject   Re: st: Absolute beginners guide to Multivariate probit analysis using Stata 12
Date   Tue, 13 Nov 2012 15:06:00 -0500

On Tue, Nov 13, 2012 at 2:58 PM, Sharon Brody <> wrote:
> I am new to Stata and new to statistics.
> I have a dataset from a survey of 12200 people that I need to analyse. I have worked out binary probit and ordered probit, but am struggling with Multivariate probit.
> I have 5 variables that I need to assess there correlation with each other but more importantly the relationships between these and the other variables in the dataset. (Simultaneous equations I think its called)
> These variables are: % food recycled, % glass recycled, % metal recycled, % paper recycled and % plastic recycled. Each variable is ranked 0 to 100%. I have put them into 5 groups of 0%, 1-25%, 26-50%, 51-75% and 76-100%. I have also created variables for Yes/No ie do you recycle - respondents saying yes are 1 and those saying no are 0. We also have the Sweights to add.

How skewed are the responses on the original scale? If they're not
very skewed, I'd skip using probit analysis or the discretization that
you have undertaken and analyze the responses using -xtmixed-, because
you have five outcomes repeated per person. It's quite likely,
however, that this is not true, in which case you are stuck with a
number of possibilities. Multilevel ordinal probit or logit can be
done using -gllamm-.

An alternative is to use ordinal logit and then pool using -suest-.
The kind of model you get is different as it's not individual level
the way that a multilevel model is, but depending on your purposes
that may be more appropriate.

Other things to think about:

What are your N's like?
How much missing data?

However, all that said, if you're new to statistics this kind of
analysis is relatively challenging, so you might want to enlist some
local talent.

*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index