Re: Logistic regression or Poisson regression (log linear)
- From: "Thom" <thom@xxxxxxxxxxxxxxx>
- Date: Tue, 9 Aug 2005 12:18:38 +0100
I've been looking at some analyses with similarly sparse data. Dichotomizing
prior to logistic regression will probably lose a lot of information. I'd
start with a very simple option such as the t test with and without
transformations such as the square root. More complex alternatives might be
ordinal logistic and maybe Poisson regression or more likely Negative
Binomial regression, but if the fit of the transformed or untransformed t
test model is good (e.g., by checking the residuals) the more complex
analyses should tell you much the same thing.
Thom
<Saxatilis@xxxxxxxxx> wrote in message
news:1123080475.093590.36560@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> Greetings,
>
> I am trying to determine which form of regression to use. The dependent
> variable is binomial, meaning either 'experimental treatment' or
> 'control treatment.' The independent variable is either 'event
> occurred' or 'event did not occur' but an event can occur more than
> once (independent events) over the course of my sample unit which is
> one day. The events are relatively rare. I have 244 samples/days. 10
> times an event occurred once in a day. Once two events occurred in a
> day, one time three events occurred in a day and one time seven events
> occurred in a day.
>
> My objective is to determine if the event occurs with the control
> treatment more often than with the experimental treatment. I am not so
> interested in determining if, for example, 7 events occur more often in
> the control treatment vs. the experimental treatment, particularly
> since this and other multiple event days happened only once. I realize
> I could bin the event data into 1 or more but this seems to be
> discarding relevant data that is already pretty slim.
>
> I hope this is clear and someone may have suggestions. Thanks for your
> considerations.
>
> Ryan Silva
.
- References:
- Logistic regression or Poisson regression (log linear)
- From: Saxatilis
- Logistic regression or Poisson regression (log linear)
- Prev by Date: Modal Divergence
- Next by Date: Re: Agreement for one subject with gold standard, ordinal data
- Previous by thread: Logistic regression or Poisson regression (log linear)
- Next by thread: Dummy Variable?
- Index(es):
Relevant Pages
|
