>Please reply to :
Suzanne Penfold <[log in to unmask]>
Good afternoon
>For my PhD analysis I have developed multiple logistic regression models
>for factors (social and psychosocial) associated with reported sexual
>intercourse in a sample of Scottish adolescents. These models are
>exploratory, i.e. I am not testing to see if a particular factor is
>associated whilst controlling for others but looking to see which factors
>are associated and what the strength and direction of each is.
>
>For several of the steps in this process I need a reference to back it up.
>Can anyone help with the following?:
>1. I developed the model in a random half of the population, tested in the
>second half and finally ran the checked model in the whole population in
>order to check that it is valid. Is there a reference for this method or a
>paper that has used this method?
>2. I have developed several different models with Cox and Snell R square
>values of about 15-20%. I think this is reasonably high for a behavioural
>model but cannot find a reference to back this up.
>3. The percentage correctly predicted by each model is about 70%. Again,
>this seems quite good but I need a reference to back me.
>
>Thanks in advance,
>Suzanne
>
>
>------------
>Suzanne Penfold
>Dugald Baird Centre
>University of Aberdeen
>Cornhill Road
>Aberdeen
>AB25 2ZL
>
>01224 553445
>
>[log in to unmask]
>
>'A Symposium on Women's Health: Measurement Matters'
>Dugald Baird Centre 10th Anniversary Symposium
>9th - 10th May 2005
>For more details contact [log in to unmask]
>or go to http://www.abdn.ac.uk/dugaldbairdcentre/
------------
Ann Fitzmaurice
Medical Statistician
Dugald Baird Centre for Research on Women's Health
Aberdeen Maternity Hospital
University of Aberdeen
Aberdeen AB25 2ZL
Uk
[log in to unmask]
Tel . 01224 553876
|