First of all: thanks and congrats for a great video! 1 kind comment: When putting categorical variables in the Factor box, you can actually 'flip' the reference group to the first one, via the "options" button. This solves your issue and is more in line with theory.
Thank you so much for your feedback! Yes, you are correct. Believe it or not, I've used that option before, but I failed to remember using it when I was putting this video together. I really do appreciate you visiting and taking time to comment! Best wishes!
Query "McFadden’s pseudo R-square" what will be the acceptable range and professor what about VIF before going for the NB model please share your thoughts
Thank you so much for these valuable videos!! Can we use Poisson regression to estimate for example the annual prevalence rate of a binomial outcome through out a 10-year period while adjusting for factor variables?
Hello there. Thank you for visiting! After doing a bit of looking online (there does not appear to be much discussion on the issue of underdispersion; most is concerned with overdispersion), I found a posting by Joseph Hilbe (see stats.stackexchange.com/questions/67385/what-is-the-appropriate-model-for-underdispersed-count-data), who recommended hurdle model or generalized Poisson regression. Neither of these options are available in SPSS unfortunately. I found this (journals.sagepub.com/doi/pdf/10.1177/1536867X1201200412) article online that refers to a an approach with Stata where you may be able to use generalized Poisson regression. It is worth mentioning that the authors of that article suggest that with underdispersion, standard errors tend to be overestimated. Logically, the effect would be deflated test statistics and loss of power. An option (if you are assuming underdispersion) when running your model using SPSS is to adopt a more liberal alpha level to offset loss of power that might arise as a result of underdispersion. However, that presumes you have correctly identified a problem with underdispersion; and liberalizing alpha might not be something reviewers would necessarily accept. I hope this helps!
This is a great video, thank you! Can I use NBR or Poisson regression in analyzing count data from a longitudinal dataset with one entity, say the number of weekly deaths from a disease as the outcome variable (and some mix of continuous and factor variables as predictors)?
![BLACK BAG - Official Trailer [HD] - Only in Theaters March 14](http://i.ytimg.com/vi/Du0Xp8WX_7I/mqdefault.jpg)
Professor Many doubts are cleared i have followed the slides and the slides are evident of your quality work expertise and hard work,Respect
Thank you so much for posting, your videos are an absolute life saver!!
You are so welcome!
Practical and informative illustration ! Thank you so much.
First of all: thanks and congrats for a great video!
1 kind comment:
When putting categorical variables in the Factor box, you can actually 'flip' the reference group to the first one, via the "options" button. This solves your issue and is more in line with theory.
Thank you so much for your feedback! Yes, you are correct. Believe it or not, I've used that option before, but I failed to remember using it when I was putting this video together. I really do appreciate you visiting and taking time to comment! Best wishes!
Very informative as usual.
Thank you so much!!!!!!!!! Very useful!!!!!!!!!
You are very welcome!
So helpful. Thank you so much!
Do you have a video where you model interactions and their interpretations?
Query "McFadden’s pseudo R-square" what will be the acceptable range and professor what about VIF before going for the NB model please share your thoughts
Thank you so much for these valuable videos!! Can we use Poisson regression to estimate for example the annual prevalence rate of a binomial outcome through out a 10-year period while adjusting for factor variables?
Hello, but how do we do quasi poisson?
It is such a helpful lecture. When there is under dispersion, can you suggest how to do it in SPSS?
Hello there. Thank you for visiting! After doing a bit of looking online (there does not appear to be much discussion on the issue of underdispersion; most is concerned with overdispersion), I found a posting by Joseph Hilbe (see stats.stackexchange.com/questions/67385/what-is-the-appropriate-model-for-underdispersed-count-data), who recommended hurdle model or generalized Poisson regression. Neither of these options are available in SPSS unfortunately. I found this (journals.sagepub.com/doi/pdf/10.1177/1536867X1201200412) article online that refers to a an approach with Stata where you may be able to use generalized Poisson regression. It is worth mentioning that the authors of that article suggest that with underdispersion, standard errors tend to be overestimated. Logically, the effect would be deflated test statistics and loss of power. An option (if you are assuming underdispersion) when running your model using SPSS is to adopt a more liberal alpha level to offset loss of power that might arise as a result of underdispersion. However, that presumes you have correctly identified a problem with underdispersion; and liberalizing alpha might not be something reviewers would necessarily accept. I hope this helps!
This is a great video, thank you! Can I use NBR or Poisson regression in analyzing count data from a longitudinal dataset with one entity, say the number of weekly deaths from a disease as the outcome variable (and some mix of continuous and factor variables as predictors)?
ถ้าผ่านก็จะดีใจนะคะความคิดเห็นของดิฉันนะคะเพราะว่าสมัครที่ไหนไม่เคยผ่านเลยค่ะ
Very informative and useful video! However the dependent variable boosted my anxiety 🥲 haha just joke
Thank you so much! So helpful.