# What is the best way to handle two peak data?

Let's suppose, this is the histogram of flu patients data. As people gain flu more on autumn to winter, and more unlikely to gain flu on summer. Therefore, the amount of patients has different gap between summer and winter.

As the picture shows, this histogram has two big peak. Find out that it's because of the seasonality(summer and winter).

Here is my question, then. I want to fit some distribution with this data without any separation.. Is there are some distribution which suppose two peak??? Or, Separating the data set to make one peak is the best way to handle this data? I want to hear many ideas to handle this data.

plus, The final goal with this data, I want to make appropriate steps of hazard. For example, If the amount of flu patients over 23,000 per day, make alarm that be careful to catch the cold.. I will divide part.. And I want to fit this data to some distribution and use that distribution score and so on.// I can't just fit normal distribution to this data, Can I?

Thanks, for any ideas and feedback..

• This seems like classic, time-series seasonality kind of issues? Oct 20, 2016 at 2:13
• @MatthewGunn acutually, it is arisen from time-series data, however, I don't need date variable to assume the specific distribution.
– lisa
Oct 20, 2016 at 2:51

look up the bimodal (Multimodal) distribution- this is perhaps one way to go about it: https://en.wikipedia.org/wiki/Multimodal_distribution

• WOW..Thanks!.. I didn't know this distribution!
– lisa
Oct 20, 2016 at 2:07
• It is perhaps possible that this is actually normally distributed but you don't have enough data (i.e. it will actually fit normally as more data points are collected), but if you believe it has two peaks (which it very well might) then the bimodal is definitely the way to go.
– Rob
Oct 20, 2016 at 3:08