Re: st: St: comparing survey pre- and post- results

[Austin Nichols <austinnichols@gmail.com>]

Andrzej Niemierko <aniemierko@partners.org> :
One method is to dichotomize the outcome and use a regression for
binary outcomes to compare pre and post:
clear
input f0 f1
3  4
18 11
31 28
6  18
1  7
end
g a=_n
reshape long f, i(a) j(post)
g agree=a>3
ta a agree
logit agree post [fw=f], r


On Mon, Nov 15, 2010 at 11:33 AM, Andrzej Niemierko
<aniemierko@partners.org> wrote:
> Dear All,
> I am trying to evaluate the effectiveness of a mentoring program at our
> institution. A random sample of mentees filled in a questionnaire before the
> start of the mentoring program. A typical question would be:
>
> Question: To what extent do you agree you personally have achieved the
> following in your career?
> Career success:
> Answers: a) too early to tell b) strongly disagree c) disagree d) neither
> agree nor disagree e) agree f) strongly agree g) not applicable.
>
> The same questionnaire was given to a different sample of mentees after the
> completion of the mentoring program. Let's say that the distribution of
> responses to the above question before and after the program is as follows:
>
> Before: a)1   b)3  c)18  d)31  e)6   f)1  g)0
> After:  a)0   b)4  c)11  d)28  e)18  f)7  g)0
>
> Assuming that the question is indeed relevant for evaluating the mentoring
> program how can I quantify the difference between the post- and the pre-
> distributions of answers? How can I test whether the difference is
> statistically significant?
>
> This branch of statistics is a bit foreign to me, and I hope someone can
> help.
>
> Thank you,
> Andrzej

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/