Closed ircwaves closed 6 years ago
Author: Bobby Braswell
Timestamp: 2017-10-21T11:18:52.412Z
Created by: rbraswell
Module anomindex now done. See a5eeff0
@bziniti please take a look when you have a chance and I'll make some test outputs.
By Ian Cooke on 2017-10-21T11:18:52 (imported from GitLab project)
Author: Bobby Braswell
Timestamp: 2017-10-23T23:24:53.432Z
Created by: bziniti
I don't know how to read line 87 ts[t] = idx[t]/<float>(t+1)
but it seems to be about calculating the probabilities.
In the rain case, would it mean every number in idx is divided by 13140+1?
By Ian Cooke on 2017-10-23T23:24:53 (imported from GitLab project)
Author: Bobby Braswell
Timestamp: 2017-10-24T01:13:38.192Z
Created by: rbraswell
I found a few bugs and the code has changed a bit. One remaining thing is that I'm not dealing with the missing values appropriately. Let me fix that and get back to you. It's pretty close.
By Ian Cooke on 2017-10-24T01:13:38 (imported from GitLab project)
Author: Bobby Braswell
Timestamp: 2017-10-27T18:03:25.330Z
Created by: rbraswell
This is module "anomindex".
Seems to work ok finally.
By Ian Cooke on 2017-10-27T18:03:25 (imported from GitLab project)
Created by: bziniti
@rbraswell can you comment on the feasible of this idea? (in other words would this be too slow?)
first transform the input data to probabilities: for empirical probabilities one could: get the order (in magnitude not time) of the data then divide by count of time points
then take probabilities as input and output a normal z score see page 1576 equations 14 - 18 for an approximation method in this paper: https://github.com/e-baumer/standard_precip/blob/master/docs/lloyd_hughes.pdf