1
$\begingroup$

I have a trained model that predicts some interesting things (like energy usage) based on the weather (temperature, humidity, etc.). I would like to run a monte carlo on the model and get a probability distribution for the coming years. I could do something like look at the average temperature (humidity, etc) every day and perturb it, but that feels like a fair amount of work, and I am wondering if there are more standard ways to do this. Maybe there are public domain weather models out there that will generate multiple runs of speculative weather?

Any and all ideas welcome...

$\endgroup$
  • 2
    $\begingroup$ How many supercomputers do you have networked? In all seriousness, weather during a course of years is no longer called "weather": it is called climate. Whose climate change model are you going to pick? $\endgroup$ – whuber May 26 '15 at 22:33
  • $\begingroup$ I am not interested in an accurate prediction, and sorry if I gave that impression. I am interested in getting some kind of probability distribution that I can base economic decisions on. And I can bring a fair amount of computation resources to bear on this problem if need be, but I don't anticipate needing that. $\endgroup$ – Mike Wise May 27 '15 at 0:23
  • $\begingroup$ But if you have no interest in whether that probability distribution is accurate, the possibilities are endless: generate any distribution you like! In all seriousness, it would be helpful to know (in a quantitative sense) just what level of accuracy you need. $\endgroup$ – whuber May 27 '15 at 13:33
  • $\begingroup$ I would like to do monte carlo runs, i.e. generate a random set of weather that is a reasonalbe guess of what could occur. Probably by sampling past weather. And then do it again. And again. Until I get a probability distribution of how much energy usage my model is predicting. I can use sampling statistics to get confidence intervals. However what I want to know is if there is a more standard way of doing this, perhaps using open source weather pediciton packages. It seems not... $\endgroup$ – Mike Wise May 27 '15 at 13:59
  • $\begingroup$ I am not stymied, and I know a way to proceed which is reasonable, and can be made more so with more work. I am just looking for alternate suggestions. Due diligence so to say. $\endgroup$ – Mike Wise May 27 '15 at 14:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.