# Is data is over dispersed?

**URL:** <https://community.drivendata.org/t/is-data-is-over-dispersed/1655>\
**Category:** DengAI Competition\
**Created:** [December 5, 2017, 8:34pm UTC](https://community.drivendata.org/t/is-data-is-over-dispersed/1655 "2017-12-05T20:34:17Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![sohaib07](https://avatars.discourse-cdn.com/v4/letter/s/d26b3c/32.png) [@sohaib07](https://community.drivendata.org/u/sohaib07)\
**Post date:** [December 5, 2017, 8:34pm UTC](https://community.drivendata.org/t/is-data-is-over-dispersed/1655/1 "2017-12-05T20:34:17Z")

</div>

Hello Everyone, I am new to Data science. I have a query is data over dispersed because it contain lots of zero’s in many column. I was going to apply negative binomial regression but for over dispersion of zeros, zero inflated regression is appropriate.

---

<div class="post-metadata">

**Author:** ![barrycv](https://avatars.discourse-cdn.com/v4/letter/b/3ab097/32.png) [@barrycv](https://community.drivendata.org/u/barrycv)\
**Post date:** [December 7, 2017, 2:35pm UTC](https://community.drivendata.org/t/is-data-is-over-dispersed/1655/2 "2017-12-07T14:35:27Z")

</div>

Over-dispersion happens when the variance\>mean. An assumption of a Poisson distribution is that those two are equal. An abundance of zeroes in your count variable can be a cause of over-dispersion. I believe most of the time analysts will use other methods when this occurs, in particular negative binomial, ZIP (zero inflated poisson), and ZINB (zero inflated negative binomial) & see which performs best.