Wednesday, March 15, 2017

interpret output, Binomial Logistic Regression, spss

https://statistics.laerd.com/spss-tutorials/binomial-logistic-regression-using-spss-statistics.php#procedure -- 頁面下方

4.12 The SPSS Logistic Regression Output
http://www.restore.ac.uk/srme/www/fac/soc/wie/research-new/srme/modules/mod4/12/index.html

log odds is different from odds ratio
說明及比較 http://zencaroline.blogspot.tw/2017/03/logit-vs-probit.html

for spss

Odds Ratio is labelled as Exp (B)

if male=0 (reference category)  female=1
The Exp(B) column (the Odds Ratio) shows 1.37.
Means that females are 1.37 times more likely than males (reference category, code 0) to achieve fiveem, after controlling for other independent variables.

Dichotomous (or dummy) explanatory variables

For a dichotomous explanatory variable the OR is simply the difference between the odds for the base category (x=0, male) and the other category (x=1, female). Thus in our earlier example for gender and aspirations the OR was 2.0 indicating girls (x=1) were twice as likely as boys (x=0) to aspire to continue in FTE.

OR is 2. While the OR is sufficient for meaningful interpretation, some researchers also like to express the OR in percentage terms. Subtracting 1 from the OR and multiplying by 100 gives the percentage change. Thus (2-1) *100 = a 100% increase in the odds.
 (1-OR) *100

If odd ratios is 1.37, females (code 1) are about 1.37 times (or 37%) more likely to achieve fiveem than males (code 0).
(1.37-1)*100= 37%

If OR is 1.48, (1.48-1)*100= 48 %

No comments: