I have fitted a GEE glm to my monthly rainfall data set and I came up with an adequate model (including month and temperature as predictors). I wanted to use this model to forecast rainfall in future years. How can I proceed?
$\begingroup$
$\endgroup$
3
-
2$\begingroup$ So, to predict rainfall in July 2019 say you need to know that it is July (easy) and what its temperature will be. I think you're stuck unless you get a temperature prediction from elsewhere. $\endgroup$– Nick CoxCommented Aug 22, 2017 at 10:11
-
$\begingroup$ Thanks @ Nick Cox. Is there any way to obtain the predicted temperature? $\endgroup$– user40494Commented Aug 22, 2017 at 10:56
-
$\begingroup$ There is an entire subdiscipline devoted to climate modelling. Downscaling to station level is a different deal however. In short, you need a different model to do that. $\endgroup$– Nick CoxCommented Aug 22, 2017 at 11:02
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
0
In general, create a new dataframe df with columns month and temperature, that represents future dates. Then you get predictions for those new records with
predict(model, df)
However, probably you don't "know" the temperature next week, so this dataframe is hard to create. Either you restrict your model to variables that you know beforehand (month), or you have some source of predictions for temperature, for example from the government, that you assume to be good, and use those.
Alternatively, off course, you can try and predict temperature yourself (only on month, or something that is known in advance), and plug these predictions into your model for predicting rainfall.
But this is very close, or in some cases even equivalent, to directly estimating rainfall on month. It can also go wrong. This all depends on what models you are using.
