It fits and it all makes sense. I've been trying to work out which one to do and how to do it for days now, playing around with the ordinal regression to no avail. This one works, im running with it. MD thesis is done. Thank you, keep up the good work and goodnight. (the regret of not having given due consideration to the power of statistics, beyond basic interpretation, over the years is painfully palpable)
Hi there, Arran! Thanks for visiting and I'm glad this was useful with your thesis! (by the way, I do have a newer video on this topic from April in case you are interested): th-cam.com/video/6S_878RheL8/w-d-xo.html) Good luck!
Thank you! It helps so much to understand and to interpret the multinomial logits models :) I do have a question: One of the assumptions is we should have independence of observations and the dependent variable should have mutually exclusive and exhaustive categories. According to my reading, it should be tested with the Haussman-Mcfadden (IIA test). But I don't think it faisable with SPSS, isn't it?
You make it sound like these parameter estimates are done pairwise against the reference category. Is that correct though? I thought the estimated probability of the reference is just 1 - the sum of all other probability, so not 1 - the one probability we are just focussing on.
Hi Mike. I really, really need your help and advice. May i share my spss problem with you? i have this warning message : "Unexpected singularities in the Hessian matrix are encountered. This indicates that either some predictor variables should be excluded or some categories should be merged. The NOMREG procedure continues despite the above warning(s). Subsequent results shown are based on the last iteration. Validity of the model fit is uncertain'. I am not sure how to troubleshoot this Dr. I am a newbie. Google, youtube, researchgate and reddit could not help me =(
thanks a lot, sir, how to adjust several variables into models and test the OR and CI .... ex (model 0ne blood pressure, age, gender. and model two contains model one + cholesterol, creatinine ,albumin )
Dear Mike Crowson. thank you for the nice and more interactive lecture. i found that, the video is more explanatory and like soft drink. what i want to ask for any one who are interesting is that' can we consider MLR for antenatal care visists??' becase WOH recommended than any pregnant mother must visit at least four times during the whole pregnancy ( Firist visit, second visit, third vist and four and above visits are the times). but to say adequate visits mothers must come four at least times during the whole pregnancy?
You said Male is 0 and Female is 1 but on the screen at the 2 minute mark, it shows genderid is 1 and 2. So which is the correct ID? Is 1 male and 2 female?
I am running multinominal logistic regression and the likelihood ratio (chi square) is insignificant, but the predictor is significant in the parameter estimates. Should i have to interpret them or i have to avoid them in my analysis?
I have two questions: 1) How can I use the AIC and BIC info to help me in my analysis/model selection in addition to the p-values that you have shown here? 2) With the results from the analysis, how can I make a prediction on someone if know his/her gender, age, econlib and religious info? Thank you in advance!!!
Hi, again Mike! I checked your video and thank you for your explanation of multinomial logistic regression (MNLR). Now the question I have is : MNLR assumes the independence of irrelevant alternatives. Most often McFadden test is used to check for that, my question is, how do you check that with STATA? Or do you just assume that this assumption holds?
Hi! i have ordinal DV an i ran Ordinal Logistic regression but it didnt met proportional odds assumption then i ran Multimoninal Logistic regression but the model fitting information is not significant. what should i do?
I have a question. If our Pearson and Deviance goodness-of-fit tests are significant (means the model is not a good fit) are we still able to use the data in the parameter estimates? Or will the data be not valid anymore? If yes, is there any solution to solve this?
Hi Mike, may I ask why you classified the categorical independent variables as covariates rather than factors? I had thought that covariates were for continuous variables and factors were for categorial variables.
Hi Erica, basically the categorical variable in this model was a binary predictor that was dummy coded. It was used in the same way that you would use a dummy coded predictor variable in the context of OLS regression. There's nothing mathematically "off" by doing it this way. If I'd treated "genderid" as a factor variable, I would've had to specified the reference/baseline category (an extra step or two) when going through the Factor route. But the results would've been exactly the same. My general preference is to use the Factors box for categorical variables that (a) have more than two categories and (b) would like the program to do the dummy coding for me. However, FYI, if chose to manually create dummy coded variables to reflect group membership on the factor variable, I could've included those in the Covariates box and accomplished the same thing. I just generally prefer to take the route with fewer steps if I already have a dummy coded variable I am using as a predictor. [Here is a video on dummy coding you might be interested in using OLS regression as an example, where I include factor variables in a model through the use of dummy coding: th-cam.com/video/XGlbGaOsV9U/w-d-xo.html . I hope this helps!
Hi there. It's been awhile since I put this presentation together, so I am not entirely sure what you are referencing. However, I have a newer presentation on multinomial LR that I put together last April that may answer your question. The TH-cam video is found at: th-cam.com/video/6S_878RheL8/w-d-xo.html I also have a Powerpoint (that accompanies the video) that provides a much deeper dive on intepretation at: drive.google.com/file/d/1nKkBGrM90yRbo7ErvWe8ZxT_qiQAdgMo/view I hope this helps!
Please see my NEW 2023 video on multinomial logistic regression with categorical outcome variables here: th-cam.com/video/UIZHNTSpED8/w-d-xo.html . You will also find downloadable resources under the video description!
![[Full Episode] The Restaurant War Thailand ศึกพ่อค้าซ่าแม่ค้าแซ่บ Episode 4 | 13 ต.ค. 67](http://i.ytimg.com/vi/h8N9tY46P-M/mqdefault.jpg)
It fits and it all makes sense. I've been trying to work out which one to do and how to do it for days now, playing around with the ordinal regression to no avail. This one works, im running with it. MD thesis is done. Thank you, keep up the good work and goodnight. (the regret of not having given due consideration to the power of statistics, beyond basic interpretation, over the years is painfully palpable)
Hi there, Arran! Thanks for visiting and I'm glad this was useful with your thesis! (by the way, I do have a newer video on this topic from April in case you are interested): th-cam.com/video/6S_878RheL8/w-d-xo.html) Good luck!
Thanks immensely. Access to the resources makes it earlier to follow.
Thanks for the video. Can't we compare non-reference outcomes? for example, "donald trump" vs. "did not vote"?
You are great! You don't know how beneficial your videos. Deeply thank you!
This is very helpful. Thank you.
thank you for nice explanation
It's an incredible video. The explanation is better than a 5-star lecturer. Thanks so much for your contribution to education.
Thank you so much! You make my grad school life way easier.
Hi Jinny, thank you for your kind remark. I'm very glad you find my videos helpful! Cheers.
Thank you! You saved my day.
Glad I could help!
Thank you! It helps so much to understand and to interpret the multinomial logits models :) I do have a question: One of the assumptions is we should have independence of observations and the dependent variable should have mutually exclusive and exhaustive categories. According to my reading, it should be tested with the Haussman-Mcfadden (IIA test). But I don't think it faisable with SPSS, isn't it?
You make it sound like these parameter estimates are done pairwise against the reference category. Is that correct though? I thought the estimated probability of the reference is just 1 - the sum of all other probability, so not 1 - the one probability we are just focussing on.
Hi Mike. I really, really need your help and advice. May i share my spss problem with you? i have this warning message : "Unexpected singularities in the Hessian matrix are encountered. This indicates that either some predictor variables should be excluded or some categories should be merged.
The NOMREG procedure continues despite the above warning(s). Subsequent results shown are based on the last iteration. Validity of the model fit is uncertain'.
I am not sure how to troubleshoot this Dr. I am a newbie. Google, youtube, researchgate and reddit could not help me =(
thanks a lot, sir, how to adjust several variables into models and test the OR and CI .... ex (model 0ne blood pressure, age, gender. and model two contains model one + cholesterol, creatinine ,albumin )
Do we check any assumptions before we do the test ?
its clearly informative vedio....thanks
Great video. Thanks--I'm running multinomial regression on parent involvement with homework. Could you update the links? (Getting a Google error)
Hi there, thanks for your post. I have just updated them. They should be working now. Cheers!
@@mikecrowson2462 Thank you
good information for my bachelor thesis, thank you so much!
Dear Mike Crowson. thank you for the nice and more interactive lecture. i found that, the video is more explanatory and like soft drink. what i want to ask for any one who are interesting is that' can we consider MLR for antenatal care visists??' becase WOH recommended than any pregnant mother must visit at least four times during the whole pregnancy ( Firist visit, second visit, third vist and four and above visits are the times). but to say adequate visits mothers must come four at least times during the whole pregnancy?
very educative. however, in this case how are you going to write the mathematical model or you need to write all the equations
great video Mike, really learned something, thank you!
You said Male is 0 and Female is 1 but on the screen at the 2 minute mark, it shows genderid is 1 and 2. So which is the correct ID? Is 1 male and 2 female?
Hi! How do we apply a Bonferroni correction to account for the family-wise error rate?
How would you write up these results in APA style? Thank you!
I am running multinominal logistic regression and the likelihood ratio (chi square) is insignificant, but the predictor is significant in the parameter estimates. Should i have to interpret them or i have to avoid them in my analysis?
I didn't understand which variables go in the factor box and which go in covariant box. Why?
Just one question, could you use the Oaxaca decomposition with this method?
I have two questions:
1) How can I use the AIC and BIC info to help me in my analysis/model selection in addition to the p-values that you have shown here?
2) With the results from the analysis, how can I make a prediction on someone if know his/her gender, age, econlib and religious info?
Thank you in advance!!!
Hi, again Mike! I checked your video and thank you for your explanation of multinomial logistic regression (MNLR). Now the question I have is :
MNLR assumes the independence of irrelevant alternatives. Most often McFadden test is used to check for that, my question is, how do you check that with STATA? Or do you just assume that this assumption holds?
Yt
Hi! i have ordinal DV an i ran Ordinal Logistic regression but it didnt met proportional odds assumption then i ran Multimoninal Logistic regression but the model fitting information is not significant. what should i do?
How I can use this for data where more than 1 response option has been chosen?
I have a question. If our Pearson and Deviance goodness-of-fit tests are significant (means the model is not a good fit) are we still able to use the data in the parameter estimates? Or will the data be not valid anymore? If yes, is there any solution to solve this?
Great video !! Congratulations! This was the best video on the topic !!!
Hi Mike, may I ask why you classified the categorical independent variables as covariates rather than factors? I had thought that covariates were for continuous variables and factors were for categorial variables.
Hi Erica, basically the categorical variable in this model was a binary predictor that was dummy coded. It was used in the same way that you would use a dummy coded predictor variable in the context of OLS regression. There's nothing mathematically "off" by doing it this way. If I'd treated "genderid" as a factor variable, I would've had to specified the reference/baseline category (an extra step or two) when going through the Factor route. But the results would've been exactly the same. My general preference is to use the Factors box for categorical variables that (a) have more than two categories and (b) would like the program to do the dummy coding for me. However, FYI, if chose to manually create dummy coded variables to reflect group membership on the factor variable, I could've included those in the Covariates box and accomplished the same thing. I just generally prefer to take the route with fewer steps if I already have a dummy coded variable I am using as a predictor. [Here is a video on dummy coding you might be interested in using OLS regression as an example, where I include factor variables in a model through the use of dummy coding: th-cam.com/video/XGlbGaOsV9U/w-d-xo.html . I hope this helps!
@@mikecrowson2462 Thank you very much, Mike! And sorry for the delay; I didn't receive a notification for your reply.
@@ericah9675 No worries. I hope you have a great day!
Thanks for this video!
If gender was significant how would you interpret that without setting a reference group?
Hi there. It's been awhile since I put this presentation together, so I am not entirely sure what you are referencing. However, I have a newer presentation on multinomial LR that I put together last April that may answer your question. The TH-cam video is found at: th-cam.com/video/6S_878RheL8/w-d-xo.html
I also have a Powerpoint (that accompanies the video) that provides a much deeper dive on intepretation at: drive.google.com/file/d/1nKkBGrM90yRbo7ErvWe8ZxT_qiQAdgMo/view
I hope this helps!
Thank you! This one helps a lot especially on the interpretation 💙
I realize Im kind of off topic but does anyone know of a good website to watch newly released tv shows online ?
@Ian Aryan Ehh atm i have been using flixportal. Just search on google for it :P -billy
@Billy Shepard thanks, signed up and it seems like a nice service =) I really appreciate it!
@Ian Aryan you are welcome =)
Please see my NEW 2023 video on multinomial logistic regression with categorical outcome variables here: th-cam.com/video/UIZHNTSpED8/w-d-xo.html . You will also find downloadable resources under the video description!
Thank you so much sir!
God bless!
NEW VIDEO ON MULTINOMIAL LOGISTIC REGRESSION WITH COVID-19 DATA DEMO: th-cam.com/video/6S_878RheL8/w-d-xo.html
Можно на русском плиз!