I have a set of data and I'm running logistic regression using s.p.s.s.
I'm getting great results but when i try to run the formula ((e^(b0+b1*x1...bi*xi))/(1+e^(bo+b1*x1....bi*xi)) on excel I'm getting bad result.
Can you help me to figure out whats going on here and why I don't get the right result on the excel? Can you show me how to do it right?
the variables are : tempaug, per, and tempaug*per. and im trying to get a binary response (safe=0 not safe=1)
now, the spss plot shows that it should look like that:
P(y) = 1/ (1+ e^-(61.599-2.249*tempaug-0.137*per+0.005*per_by_tempaug))
(its the same as the formula above)
that should give me accuracy of 81.8% with cut value of 0.5
4 files attached:
1. the set of data on spss
2. the spps logistic regression plot
3. the excel file, with the formula and the results (yellow- real results; red-calculated results)
4. the spss logistic regression plot in PDF format
You said that you wanted to use the formula :(e^(b0+b1*x1...bi*xi))/(1+e^(bo+b1*x1....bi*xi). The formula you used is mathematically equivalent, but for some reason, you weren't getting the same results in Excel as you did in SPSS. To be honest, I don't know why that is, but I suspect that there was some rounding issues: numbers could have been dropped as calculations moved from columns P to Q to R ...