Timeline for Running regression in each sample separately?
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 20, 2020 at 12:00 | history | tweeted | twitter.com/StackGIS/status/1307650725892288513 | ||
| Sep 4, 2020 at 1:59 | history | reopened | Aaron♦ | ||
| Jan 5, 2019 at 4:13 | history | closed | PolyGeo♦ | Needs more focus | |
| Jan 5, 2019 at 4:10 | history | edited | PolyGeo♦ | CC BY-SA 4.0 |
deleted 43 characters in body; edited title
|
| S Nov 7, 2018 at 20:47 | history | suggested | lambertj | CC BY-SA 4.0 |
edited typos, grammar
|
| Nov 7, 2018 at 19:45 | review | Suggested edits | |||
| S Nov 7, 2018 at 20:47 | |||||
| Mar 13, 2018 at 16:08 | comment | added | Synet | Joffrey Is it work between to raster, for example, I have a Landsat scene, in this scene I calculate de NDVI and for the same scene i have de biomass. Can I use de NVDI to predict the biomass and create a R² and RMSE images? | |
| Apr 28, 2014 at 15:45 | comment | added | Jeffrey Evans | If you want to fit a global regression you will certainly need a much larger sample. You cannot possible be representing the spatial variability with 150 samples, even with a stratified random design. A 1% subsample ((12618*4144)*0.01)=522890 would be a good target n. It is critical to look at both exploratory analysis and model fit when approaching these types of problems. In your exploratory analysis you should check the sample distribution against your population. And, no you should not be producing 150 regression equations! Perhaps it is time to talk to a stats person. | |
| Apr 28, 2014 at 15:13 | answer | added | Jeffrey Evans | timeline score: 4 | |
| Apr 28, 2014 at 6:51 | comment | added | radouxju | I would say one regression per land cover, not per point, otherwise you will not be able to generalize | |
| Apr 28, 2014 at 4:10 | history | asked | Bandrush Barda | CC BY-SA 3.0 |