Re: st: Binary and ologit

[David Hoaglin <dchoaglin@gmail.com>]

Dear Aminu,

Those who are interested in giving suggestions would find it helpful
to have more information on the goals of your analysis.

So far, it seems to me that you have not described a meaningful
outcome variable for the cattle.  I thought it would be a dichotomous
outcome, diseased or not diseased; but all the cattle in your dataset
are diseased, and an analysis would not have the necessary information
to separate them into acute and chronic.

For the sheep, on the other hand, you do have a dichotomous outcome
(acute or chronic).  Since it has only two levels, you may think of it
as ordered or as unordered, whichever you prefer.

At the risk of making a silly suggestion, what would happen if you fit
a logistic regression model to the data on the sheep, and then applied
it to the data on the cattle (perhaps with adjustments in the values
of some predictor variables, to reflect well-understood differences
between the species)?  Of course, the data on clinical stage are
missing for the cattle, but (if the sheep model has a fairly high
c-statistic) the two subsets might differ in interesting and
interpretable ways.

David Hoaglin

On Tue, Jul 31, 2012 at 8:37 PM, Shittu, Aminu <ameen_vet@yahoo.com> wrote:
>  Hi Maarten,
>
> Apologies my initial posting might be vague. I am suppose to add that none of the over 12,000 animals in my 30 years dataset is healthy. We selected animals that were all diagnosed with our particular disease of interest, whose 2 clinical stages (acute and chronic) were recorded in sheep, but it only said 'diseased' in cattle - same disease though. Could this diagnosis with different levels (acute, chronic and diseased) be treated as ordinal even though cattle has only 1 level in the dataset, and 2 levels for sheep?

*
*   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/