[R] Logistic regression: categorical predictor and continuous response

[Prof Brian Ripley]

You can a logistic non-linear regresion with a continuous response -- see 
SSlogis in package stats (using nls).

However, I think

> I believe I should run a logistic regression on this

is the problem.  You have six groups of multivariate observations and you
want to know if and how they differ by group(?)  If so the usual
approaches are

- linear and quadratic discriminant analysis, very sensitive to joint 
normality.  (lda and qda in package MASS),

- logistic discrimination.  As you have > 2 groups, use multinom in 
package nnet.

If you were here I would be directing you to our statistics consulting 
service -- do you have a local source of statistics help?


On Tue, 22 Jun 2004, Christine Parent wrote:

> Hi all,
> 
> I am fairly new to R, and not a stats expert. I am having trouble finding
> information to help me analyzing my set of data.
> 
> I got morphological data on land snail shells of different species from the
> Galapagos Islands. Each island has up to 6 different vegetation zones, and
> snails have adapted to up to 4 of these zones on each of the major islands.
> 
> I would like to test if there is any correlation between the morphology of
> the shells which is continuous (I ran a PCA analysis on the morphological
> data, and the variation can be parted into two main principal components:
> PC1 for size and PC2 for shape) and the vegetation zone where these snails
> are found (which is a categorical predictor variable).
> 
> I believe I should run a logistic regression on this, but all I could find
> in R was an analysis where the predictor was continuous and the response was
> categorical. Could anyone direct me to the relevant ref or how to write the
> proper script to run a logistic regression analysis where the predictor
> would be categorical and the response continuous?
> 
> Any help or advice on this would be greatly appreciated!
> 
> Many thanks,
> 
> Christine
>   
> Christine Parent
> Ph.D. candidate
> Department of Biological Sciences
> Simon Fraser University
> cparent at sfu.ca

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595