r.series: wrong usage of weights

[mmetz]

A weighted average is calculated with

avg = sum(w[i] * x[i]) / sum(w[i])

r.series calculates weighted averages with

avg = sum(w[i] * x[i]) / count

r.series should instead behave like r.neighbors and use the corresponding weighing functions from libstats if available.

Test commands:

# set the region
g.region -p n=10 s=0 w=0 e=10 res=1
# create identical input maps
r.mapcalc "input1 = 1"
r.mapcalc "input2 = 1"
r.mapcalc "input3 = 1"
r.mapcalc "input4 = 1"
r.mapcalc "input5 = 1"

# average
r.series in=input1,input2,input3,input4,input5 method=average out=avg
# weighted average
r.series in=input1,input2,input3,input4,input5 method=average out=avg_w weights=0.1,0.2,0.3,0.2,0.1

Without weights the result is 1, with weights the result is 0.18 instead of 1. Weights have been added to r.series in r49946. I am going to change r.series in trunk and relbr70.

[mmetz]

Fixed in r68124,5 (trunk, relbr70).

[marisn]

A second candidate for erratum as the wrong result might slip unnoticed for not so trivial cases?

Affected releases: 7.0.0-7.0.3