Bookmark and Share

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

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

Re: st: xmelogit error message

From   Nick Cox <>
Subject   Re: st: xmelogit error message
Date   Mon, 22 Oct 2012 16:08:19 +0100

I can't usefully add to what I said earlier. I think you need to show
more results, e.g. all the -xtmelogit- output, before anyone has a
chance to give good specific advice.

What you are trying to fit exactly? How many parameters are you
estimating? What sample size? What frequencies for these categorical
variables? Are some combinations especially rare?


On Mon, Oct 22, 2012 at 3:56 PM, KOTa <> wrote:
> Hi Nick,
> probably i didnt explain myself clearly.
> "simplify it first and then complicate it gradually to see what is problematic."
> this is exactly what i do, the moment i add to the model any
> categorical variable with more than 2 levels - error message appears.
> all other variables are ok
> On Mon, Oct 22, 2012 at 1:54 PM, Nick Cox <> wrote:
>> -xtmelogit- is finding difficulty fitting your model. The most obvious
>> thing to do is to simplify it first and then complicate it gradually
>> to see what is problematic.
>> Nick
>> On Mon, Oct 22, 2012 at 12:29 PM, KOTa <> wrote:
>>> i am using xmelogit to estimate the model where some IV are categorical.
>>> when i am trying to use IV which are categorical predictors with more
>>> then 2 levels (i.e. not just 0/1 - those are added without problems)
>>> e.g. i.status , where status values are A to F (encoded as numbers)
>>> i receive a following message:
>>> "numerical derivatives are approximate
>>> flat or discontinuous region encountered"
>>> what does it mean?
>>> i dont use those variables correctly?
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index