In what situation is logistic regression better than the usual type of regression model? How do we interpret the coefficients of a logistic regression?
Hi Phil, is there a textbook reference of converting odds ratios to percentages? I need to be able to reference the method for my thesis. Many thanks in advance.
A really useful explanation - thank you for posting! Does the DV have to be coded 0 and 1 or can it be 1 & 2? As in, since it's nominal, does the actual number used for coding it make a difference?
Hi Phil, I've done a binary logistic regression analysis in SPSS and the outcome of goodness of fit (hosmer and lemeshow test) is significant (p.=000). Is this very problematic or can I accept this?
thanks a lot for video :D I'v performed a binary logitic regression analysis with 2 independent (categorical) variables, the p valuse were significant but OR ratio were less then 1, what migth be the reason ? what does it mean please help ((
Hi. You are talking about the Wald test - test for individual significance for each coeff. Yes, if not sig. from zero, then that regressor isn't linked to the probability of an event. However, interpreting the p-value is not as informative as looking at the confidence interval of the parameter estimates. In some journals (eg in health sciences) you'll see confidence intervals reported as well as p-values.
B is the unstandardized (regular) coefficient, and you take Exp(B) to get an interpretation of it in terms of odds ratio. I haven't seen applications of standardizing B. Indeed it seems out of the big stats packages, only SPSS gives standardized coeffcients by default. There are reasons for not using standardized coefficients - see for example, John Fox's Applied Regression Analysis.
+Emily Joa Hi Emily. Interpreting the coefficient of a nominal (categorical) variable will be in the same manner it's done for standard regression - comparison of given group to the reference group.