Poisson Regression: Zero Inflation (Excessive Zeros)

This video has been created quickly to cover in class lectures due to school closures associated with COVID-19.

This video discusses what zero inflation (or excessive zeros) is in Poisson regression, how to detect zero inflation, the common causes of zero inflation, and solutions to address zero inflation. Specifically, we mention zero inflated poisson (ZIP) models, and hurdle regression.

►► Watch More:


Follow MarinStatsLectures


Our Team:
Content Creator: Mike Marin (B.Sc., MSc.) Senior Instructor at UBC.
Producer and Creative Manager: Ladan Hamadani (B.Sc., BA., MPH)

These videos are created by #marinstatslectures to support some statistics courses at the University of British Columbia (UBC) (#IntroductoryStatistics and #RVideoTutorials ), although we make all videos available to the everyone everywhere for free.

Thanks for watching! Have fun and remember that statistics is almost as beautiful as a unicorn!

#statistics #rprogramming

[ashtondickerson]

Thanks so much for your Poisson series! So helpful

[soumitrakar]

Can you kindly add the Overdispersion and excessive zeroes and two Part videos of Poisson regression of Mary Clare Kennedy also a part of the video lecture series of Poisson regression then it will be helpful for viewers.

[knosis]

Thanks for the great definition. You are a life saver!

[haowu]

well explained the zero-inflated models, thanks lot!

[RqueErre]

Many thanks, it's been very clear and helpful!

[marieguittonneau]

Thank you ! Do you know if there is an adaptation of Generalized Linear Models that allow the data to follow a zero-inflated Poisson distribution ?

[meleksenakn]

Hi Mike,
Could you provide the R scripts which you have mentioned on the videos of Poisson Regression Part 2 and Part 3?

[梁文-ri]

Very interesting, what happen if zeros are less than expected, extending the patient visits example, if there are a group of people refused to see doctors because of some reasons.

[EvanZamir]

How do you predict "Always Zero"?

[CPO_]

Thanks again for this nice lecture series!

[SonuSingh-dlcj]

Hi I am trying to perform a regression where loan amounts are the dependent variable however as in such cases there are many 0's I was looking for an appropriate model. Although the data is not count data i.e Many people could receive $0, However nobody will receive $1, $2, could I still use the Zero Inflated Poisson Model? I understand I can do a binomial model however this will not explain the features which lead to high amounts just which lead to loans