Calculating predicted probabilities from a logistic regression model SPSS

#1
Hello everyone,
I am trying to calculate individual predicted probabilities from a logistic regression model with SPSS (to describe how a individual probability could be calculated from my model in an article). I was looking for a formula how the predicted probabilities in SPSS are calculated (automatically when you check "save predicted probabilities"). I've found this video https://www.youtube.com/watch?v=6jj4muvWDgs , which explains quite clearly how the probabilities are calculated in SPSS: 1st step - predicted_logit calculation; 2nd step - predicted probability calculation. This approach works for me as well if I go this way, but in other sources, i.e. here http://www.pmean.com/13/predicted.html the probabilities are calculated differently and yield different individual probabilities.

Therefore my question: is the approach described in the first video correct, and if it is, is it possible to combine the two steps into one relatively simple equation to use in an article?

Thanks in advance!
 

rogojel

TS Contributor
#2
hi,
without looking at the links, the equation you look for is:

ln(p/(1-p))=a0+a1*x

Expressing p is then just a bit of manipulation.

regards
 
#3
Thanks! p stands for probability, a0 for the coefficient on the constant term, a1 for the coefficient(s) on the independent variable(s) and x for the independent variable(s)?
 

rogojel

TS Contributor
#4
hi,
Yes. You can see the full equation in the output, in case of more variables, interactions etc... it is the same idea

regards
 

gianmarco

TS Contributor
#7
Hello,
the formula is right after the first paragraph in the "Materials and Methods" section.
I have found this YT video that can also help you in calculating the predicted probability:


[YOUTUBE]eX2sY2La4Ew[/YOUTUBE]

cheers
gm