Hi there,

Does anyone know if its OK to perform logistic regression with only one independent variable predicting the dependent variable?

Id be super grateful for any advice.

:-)

hullo there!! I'm a bit thick when it comes to statistics but I have a very helpful statistics book in my office which i will check tomorrow and give you an actual answer, thought I'd stop in now and let you know cos I didn't want to leave you hanging!!

I'm not an expert, but...what do you mean by "OK" ? Are you able to give any more details about what it is you are modelling the relationship between?

Hi both,

Thanks so much for your replies. Im interested in finding out if cognitive function (continuous IV) predicts work status (working / not working)

Thank youuuuu!!!!

Hello Buttercup1984, would you like a simple introductory book on logistic regression, Logistic Regression, 2nd Edition? If so PM me your e-mail address and I will send it to you. Oh, and as far as I'm aware, you can use just one independent variable to predict one dependent variable :-)

Illegal!

AFAIK, yes you can model this relationship in this way, but the validity of your model depends on a lot of things (sample size and so on). Probably the book that walminskipeas mentioned will cover checking the fit of your model - such things are covered in the book I have, which I would equally offer to send you, only it does not belong to me but to the library. Most of the contents of the book is on Google Books though:

http://books.google.co.uk/books?id=Po0RLQ7USIMC&lpg=PP1&dq=applied%20logistic%20regression&pg=PP1#v=onepage&q=&f=false

believe it or not I teach logistic regression to postgrads HA!

I recommend Andy Field - his book is fab, and he also has a really good website that have his lecture hand outs on.

http://www.statisticshell.com/cocytus.html

ya! one variable should be ok! I tried andy fields book but even that was too much for my relatively limited brain, i don't understand why I can't get my head around stats...anywhoo, the book i use is Choosing and Using Statistics, A Biologist's Guide, by Calvin Dytham, it's a bit basic but is thorough enough to make sure you understand what test you need and that youre doing the right one and shows you how to do it in spss :)

Wow thank you everyone your advice has been great!!

I'll look into everything everyone said and send some pm's too.

THANK YOU!!

Buttercup:-)

While it is ok to find the relationship between a dependent variable and one independent variable, I would think this is too simple a model. I would suggest controlling for other factors as well. You might find the relationship is not so stong after all given other factors and this is what real world data is like. My suggestion is 1) run the simple model first, and 2) try controlling for other factors (if there are other actors) and see what happens. I hope this helps.

good idea. I'll run the simple model first to 'find my feet' and then see what I can control for. Thanks so much for your advice I really appreciate it.

By the way Buttercup, Roger Tarling is the best on regression. Andy Fields as well.