search

USGS-R / protoloads

Prototyping and exploring options for broad-scale load forecasting
0 stars 4 forks source link

fig: predicted vs observed by model type #57

Closed aappling-usgs closed 6 years ago

aappling-usgs commented 6 years ago

img_20180510_162024642

probably pick 1 or two lead times and just stick with those? or color by lead time? main focus should be model type

jzwart commented 6 years ago

I can start working on this

jzwart commented 6 years ago

This is with just a lead time of 0 or 1 days and y and x-axes are the same. dotted line is 1:1 line. I can change the color scheme but added a .yml so that this can be easily changed from fig_config.yml: image

aappling-usgs commented 6 years ago

Whooooeeey, site number 3! That's some crazy autocorrelated error!

It's too bad all the points are compressed for site 2, but otherwise I think this plot shows exactly the info I'd want. Opens up discussions about bias and autocorrelation but also makes it clear that the predictions can be pretty decent when conditions are right :+1: