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]

st: ST: re: logistic regression and 'adequate numbers'

From   Ronald McDowell <>
To   "" <>
Subject   st: ST: re: logistic regression and 'adequate numbers'
Date   Thu, 7 Apr 2011 13:12:38 +0000

That's helpful Martin.

The slight complication is that my data is survey data and has probability weights and is stratified, so I can't get standardised residuals from STATA . Im told this is the case with other packages, though am happy to be proved wrong. However, if formulae were available for the variance-covariance matrix of the residuals under these conditions, I could attempt to compute my own. There's always the possibility of looking at unstandardised residuals, which I would have reservations about.


Date: Wed, 6 Apr 2011 09:06:29 +0100 (BST)
From: Maarten buis <>
Subject: Re: st: ST: logistic regression and 'adequate numbers'

- --- On Wed, 6/4/11, Ronald McDowell wrote:
> I'm modelling the log odds of a
> binary outcome as function of several covariates including
> age. Unsurprisingly as my sample gets older I have fewer
> numbers samppled for each year of age, and was wondering if
> anyone had come across or used a 'rule of thumb' in terms of
> how many sampled per age was 'too few' to include in the
> model? I've read of studies restricted to certain ages
> because of 'inadequate numbers' but haven't found a
> definition of the latter.
> I'm aware I can perform sensitivity analysis, look at
> residuals and the usual diagnostics (not all of which are
> available in STATA in this case unfortunately), but if
> anyone had any thoughts on the matter I'd be interested.

There is a fairly complete set of types of residuals and other
diagnostic tools after you estimate your model with -glm-.
When it comes to age, I would not be worried about the number
of observations but about selective mortality: people who
survive to a very old age are different from the general
population in many ways.

Hope this helps,

- --------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
- --------------------------

*   For searches and help try:

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