I did u multivariate logistic regression analysis, so I have dummyvariables, but I want to know if the dummyvariables as a set together are significant to my dependent variable. How can I do that?
You would do a Wald test for the significance of all the coefficients. See my video on testing hypotheses after regression for an example. Another way is to leave out the dummy variable and do a likelihood ratio test between original model and the simpler model. See my video on model comparisons.
@@mronkko Thank you for your reply. Is it also possible to examine it with a R2 test? So I know if the set dummy variables is changing the R2? If yes, how can I do such a test with SPSS?
@@user-vu6em1jz9c I am not aware of any tests based on pseudo-R2 comparisons. I guess you could develop such test because many pseudo-R2 statistics are calculated using the likelihood statistic. But it would be more straightforward to do a test on the likelihood statistics directly. I do not use SPSS so I cannot be of much help. If I had to do a test in SPSS, I would either search the documentation for how to do a model comparison using likelihood ratio tests. If it is not possible, I would just put the two likelihood statistics into excel and do the likelihood ratio test there.
@@mronkko in order to compare models? I have a single model to each variable, then I wish to compare them to see what's the best one, but I dont know if i can rely on the pseudo R
@@larissacury7714 Best in terms of what? R2 quantifies three different aspects of model fit and Pseudo-R2's at most two. The choice of a statistic depends on what aspect you are interested in.
@@mronkko Well, I have one Dependent variable but I've made 3 single models, one with each indepent variable. I'll put them all on the same model later on, but, for now, I'd like to compare them independently
Thanks for the video but I have question for two legit models one of them is restricted and other one is unrestricted. How can we find which model fits the data better?
Depends on what exactly you mean by model fit. The pseudo-R2s quantify model fit in different ways and you should choose the one that quantifies the aspect of model fit that you are interested in. If the models are nested, the unrestricted model fits always better or equally well than the restricted one. If you want to test the statistical significance of the difference in model fit (in the likelihood sense) you can use likelihood ration test for nested models.
Thank you for the help! Will check the rest of your content
I must warn you that there is a lot to check ;)
Thanks for motivating why we should not report Nagelkerke R-Square
You are welcome.
Thank you for your explanation Mikko!
Thank you for the clear video. This is very helpful.
Thanks for such an amazing video
You are welcome!
What is the best R2 measure for multinominal logistical regression?
Depends on what you want to quantify with the pseudo-R2.
I did u multivariate logistic regression analysis, so I have dummyvariables, but I want to know if the dummyvariables as a set together are significant to my dependent variable. How can I do that?
You would do a Wald test for the significance of all the coefficients. See my video on testing hypotheses after regression for an example. Another way is to leave out the dummy variable and do a likelihood ratio test between original model and the simpler model. See my video on model comparisons.
@@mronkko Thank you for your reply. Is it also possible to examine it with a R2 test? So I know if the set dummy variables is changing the R2? If yes, how can I do such a test with SPSS?
@@user-vu6em1jz9c I am not aware of any tests based on pseudo-R2 comparisons. I guess you could develop such test because many pseudo-R2 statistics are calculated using the likelihood statistic. But it would be more straightforward to do a test on the likelihood statistics directly.
I do not use SPSS so I cannot be of much help. If I had to do a test in SPSS, I would either search the documentation for how to do a model comparison using likelihood ratio tests. If it is not possible, I would just put the two likelihood statistics into excel and do the likelihood ratio test there.
I really respect you....
You are welcome!
Thank you! Very helpful indeed!
Glad it was helpful!
If the Negel R² is not reliable, then what would be the best way to compare ordinal logistic regression models?
A Negel R² of 0.3, then, would not be that bad?
The answer depends on what is the purpose of the comparison. The different R2s quantify different aspects of model fit.
@@mronkko in order to compare models? I have a single model to each variable, then I wish to compare them to see what's the best one, but I dont know if i can rely on the pseudo R
@@larissacury7714 Best in terms of what? R2 quantifies three different aspects of model fit and Pseudo-R2's at most two. The choice of a statistic depends on what aspect you are interested in.
@@mronkko Well, I have one Dependent variable but I've made 3 single models, one with each indepent variable. I'll put them all on the same model later on, but, for now, I'd like to compare them independently
It is good but can,t down load
RUclips does not allow downloads of videos. But why would you want to even do that given that you can just access it from the website?
Thanks for the video but I have question for two legit models one of them is restricted and other one is unrestricted. How can we find which model fits the data better?
Depends on what exactly you mean by model fit. The pseudo-R2s quantify model fit in different ways and you should choose the one that quantifies the aspect of model fit that you are interested in. If the models are nested, the unrestricted model fits always better or equally well than the restricted one. If you want to test the statistical significance of the difference in model fit (in the likelihood sense) you can use likelihood ration test for nested models.
well explained thank you.
You are welcome