Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Test selection for effects of a number of variables on a dependent variable

 From Nick Cox To "statalist@hsphsun2.harvard.edu" Subject Re: st: Test selection for effects of a number of variables on a dependent variable Date Wed, 12 Jun 2013 11:49:38 +0100

```Several confusions here.

Quite what you mean by "factor variable" is unclear (I surmise that
you may be confusing Stata with R, for example), but, regardless,
-regress- allows any kind of numeric variable to be the response, even
down to the level of a binary variable. How far that is a good idea
depends on the circumstances and Stata trusts you to make a sensible
decision.

There is no limitation whereby -mlogit- is restricted to 3 outcomes.
What prompted that? Numerous different outcomes may be difficult or
even impossible to model comfortably or at all, but that's a different
matter.

What's crucial is exactly what "stage of diagnosis" means. It could,
for all I know, a graded or ordinal scale, in which case I would try
-ologit- first.

I have to recommend _much_ more reading before analysis. These are
basic misunderstandings of statistics and indeed Stata. The textbooks
of Alan Agresti are excellent choices.

Nick
njcoxstata@gmail.com

On 12 June 2013 11:22, Tim Evans <Tim.Evans@wmciu.nhs.uk> wrote:
> Hi all,
>
> I am using Stata 11.2, and have a cohort of patients diagnosed with a particular type of cancer and have a couple of questions to ask of my data. The first relates to the effects of certain variables (social isolation, deprivation, stage at diagnosis etc) on survival time - this I think I have sorted out. However, one question I have to answer is:
>  'does ethnicity or social isolation have an effect on stage at diagnosis (controlling for the other factors).'
>
> My stage at diagnosis variable has 6 different values, social isolation has 4 different variables and ethnicity has 5 variables. Originally I thought I would be able to model this using -mlogit-, but this appears to me, to only be appropriate when the dependent variable has 3 different values (and mine has more). I naively tried to use -regress- but the dependent variable cannot be a factor variable and the data I have are not categorised in a continuous manner.
>
> Therefore, does anyone have any suggestions on how I might be able to model the effect of one of my variables (ethnicity or social deprivation) on stage at diagnosis?
>
>
> Best wishes
>
> Tim
>
> _DISCLAIMER:
> This email and any files transmitted with it are confidential. If you are not the intended recipient, any reading, printing, storage, disclosure, copying or any other action taken in respect of this e-mail is prohibited and may be unlawful. If you are not the intended recipient, please notify the sender immediately by using the reply function and then permanently delete what you have received.
>
> The information contained in this e-mail may be subject to public disclosure under the Freedom of Information Act 2000. The confidentiality of this e-mail and your reply cannot be guaranteed, unless the information is legally exempt from disclosure.
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

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