# Thread: Combining multiple probabilistic weather signals with cumulative binary dist

1. ## Combining multiple probabilistic weather signals with cumulative binary dist

If I have a stable of separate, independent models (N = 19 of them) that are using different inputs to each generate a probabilistic forecast for rain (x, or success) tomorrow, what is the best way to aggregate these probabilities into an overall/grand probability given the stable of models' percentage of successes?

I've been working with the cumulative binary distribution function in excel and matlab, and have experienced counter-intuitive results. For example, given N = 19 and x = 10, and probability of success (the average probability from each of the 10 models that are firing/predicting rain tomorrow) of 60%, I return a 33.3% cumulative binary distribution probability. If I lower probability of success to 58% (keeping N and x same), I get a 40% cum bin dist. Lower p, higher cum bin dist, etc. How do higher probabilities of success with same N and x lead to lower cum binary dists, or the area between 10 and 19??

I must not be interpreting the results right, or am missing something very important here...which are both euphemisms for being totally wrong here (I think). My goal is to roll the discrete success probabilities up into an aggregate probability (hopefully) getting "lift" from a larger majority of success models calling for rain.

Thank you

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts