#### SPSS

B

hi guys,
im currently analyzing my data using SPSS and generally regression,

has anyone done this,? I could use some tips...

thanks!

D

I have done regression analysis for my data with SPSS, don't profess to know much but what are you after?

B

thanks!

well, a simple question, i'm confused regarding confidence intervals...i'm running a logistic regression but I'm not sure what is the significance of the interval to my coefficient, what if the interval contains 1 in it?

D

When I read back on my notes for logistic regression the confidence interval was left blank.  I too am not sure what this would mean for logistic analysis. I tended to stick with linear regression for my data which I find easier to manage :\$  Sorry not a specific answer. Do you have access to statisticians at your uni to ask, this is where most of my help came from.

K

Hey Billy, I've just spent most of my day doing multiple regressions on SPSS. I am using linear regressions, and I have to confess I don't know much about logistic regression, but I would definitely advise seeing a statistician if you're not sure. I've had a meeting with our statistician today and even though I know how to do regression there are all sorts of issues that have cropped up wth my data that I hadn't really thought about before. Now I'm having to re-run some of the regressions and am in my last month of writing up- I would definitely get your stats checked out earlier rather than later so you don't end up having to change things at last minute! Best, KB

H

Quote From billy8181:

thanks!

well, a simple question, i'm confused regarding confidence intervals...i'm running a logistic regression but I'm not sure what is the significance of the interval to my coefficient, what if the interval contains 1 in it?

Are you calculating odds ratios? If the confidence interval includes 1 then it suggests that the effect you have observed (i.e. the odds ratio) is not statistically significant. You should always report an effect with a confidence interval plus some kind of statistical test (e.g. a Chi squared test which gives a p value).

B

I'm calculating both a logistic regression and a linear regression, depending on the dependent variable.....

almost all intervals contain 1 if i'm not wrong, is that possible? surely not all relationship must be rejected..!

B

my sample is 103 (out of 700) is that not large enough? must be !

H

Whether or not that is a big enough sample size depends on what you're measuring and how much of a difference you're hoping to find between the (controls and experimental) groups.

It's hard to give you exact advice via a forum. I suggest that you try to talk to a statistician in your department who will be able to look at the whole picture with you.

R

There is a video that demonstrates how to perform and interpret a logistic regression in SPSS on youtube: