## Student in dire need of help - Multinomial logistic regressi

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

lapive
Posts: 1
Joined: Sat Dec 22, 2012 10:40 am

### Student in dire need of help - Multinomial logistic regressi

Hello everybody!

I am a student in biology working on my master thesis. I am facing a huge problem, the statistical analysis.
The due date is in 1 month (a bit less) and I am not sure of what I got and this is why I turn to you today.

I used SPSS 20 to run my multinomial logistic regression on my data. So here are the screenshot of the data :

Where Temp is the temperature, Behavior is a 4 choices variable, Valley is a 2 choices variable, Month and Year are when the particular behavior was recorded. Lastly for the individual, the number correspond to the tag (so to a particual individual, I have 12 different individual).

for the non french speaker :

in type : Chaine is a string
numérique : numeric
Aucun = none
If you have other doubt just ask!

So now the important part, here is what I ran :

And here is what I got :

Here is the output of SPSS :

Now my questions/concerns :

I have, for all my effects, a 0.000 p-value and it seems odd (particulary, I wasn't expecting a p-value < 0.05 for the valley) so here is my main concern, is my data formated in a good way ? do I need to enter some specific value somewhere in the options?

and my second question, how can I put those results in a graph? I have no idea and couldn't figure how to make a graph with a string as main axis in SPSS.

I really hope some of you here will be able to help me as I run out of time.

Thanks also for the one that will take some time for me,

Happy holidays

Lapive
statman
Posts: 2750
Joined: Tue Jun 12, 2007 12:08 pm
Location: Florida, USA

### Re: Student in dire need of help - Multinomial logistic regr

Initially, a suggestion just to make sure SPSS is doing what you think - Do an auto recode on the string covariate then use that in the model

Did you do any correlation checks upfront, these might help
See the note below