Calculations fail on years with no flow

[wknoben]

Description
This is an arid basin with long periods of no flow. In such cases the standard deviation of certain years is 0 and the correlation for such years is undefined. As a result, all metrics are calculated (partly) on NaN values and return NA.

I don't know if any clear solutions exist to make the computations succeed in these cases (maybe just hard-code solutions for the specific cases where obs and/or sim are all zero, or try to detect these and exit with a specific message?), but it seemed good to document this behaviour at least.

Sample data: gumboot_issue.csv

Sample code

df <- read.csv("gumboot_issue.csv")
bootjack(df)

Output

  GOF_stat seJack seBoot p05 p50 p95 score biasJack biasBoot seJab
1              NA     NA  NA  NA  NA    NA       NA       NA    NA
Warning messages:
1: In bootjack(df) : NAs introduced by coercion
2: In bootjack(df) : NAs introduced by coercion
3: In bootjack(df) : NAs introduced by coercion

[KevinShook]

Thanks for finding this. I'll get to it when I can, but won't be able to do anything until June.
It might be possible to use na.rm = TRUE for some functions, but it might not work for everything.

[wknoben]

Cool :) It only happens in 4 cases for what I'm currently doing but it seemed good to document this in case anyone else comes looking.