Get Logistic regressions odds from equation result

Question

By using logistic regression in SPSS I get an equation that looks like this:

 0.1535 * X1+ 
 -0.0002973 * X2 + 
 -114.9 * X3 + 
 -114.0 * X4

How do I transform the result of the equation to the odds of the classifier (between 0 and 1) using SQL?

Thanks in advance!

Learner · Accepted Answer · 2015-06-17 10:14:32Z

2

You need to use the EXP() to convert the logs odds to odds.

Example: EXP(-1.30) = 0.27 then EXP(logs odds)/(1+EXP(logs odds))

i.e.

Probability = (0.27)/(1+0.27) = 0.21

To see how to use EXP in SQL see here

Hope this helps !

answered Jun 17, 2015 at 10:14

Learner

1,5453 gold badges19 silver badges37 bronze badges

$\begingroup$ The following equation works for some of the values: Probability=EXP(Equation Result)/(1+EXP(Equation Result)) But for values of EXP(Equation Result) that are greater than 1 the method doesn't work, do you have am idea why $\endgroup$
– Rob
Commented Jun 17, 2015 at 11:18
$\begingroup$ That should work. Can you please give me an example and what error are you getting? $\endgroup$
– Learner
Commented Jun 17, 2015 at 11:22
$\begingroup$ Sure: EXP(-1.062)/(1+EXP(-1.062)) = 0.74 --> works fine EXP(0.2215)/(1+EXP(0.2215)) = 0.44 --> wrong odds --> should be 0.55 EXP(4.9)/(1+EXP(4.9)) = 0.17 --> wrong odds --> should be 0.83 $\endgroup$
– Rob
Commented Jun 17, 2015 at 11:26
$\begingroup$ hmmm.. that's strange.. The formula does not change for each row. So there is nothing wrong with formula. I doubt something wrong with the data type. $\endgroup$
– Learner
Commented Jun 17, 2015 at 11:32
1

$\begingroup$ No need to change the formula. Formula in the answer should give you the correct results irrespective of the signs of the value. see this pmean.com/13/predicted.html $\endgroup$
– Learner
Commented Jun 17, 2015 at 11:51

| Show 1 more comment

1 Answer 1