Recreation Coefficients - Missing Some Predictors

Hi, I am running the InVEST recreation model using three predictor variables. The model runs fine, but when I look at the regression_coefficients.txt there are only coefficients for one of the three variables. Am I doing something wrong?

I have attached my log and regression_coefficients text tiles.

Thank you!

Sarah InVEST-Recreation-Model-log-2020-08-05–09_48_16.txt (5.7 KB) regression_coefficients_04.txt (463 Bytes)

Thanks for posting, @sarahhalperin. This is interesting. What does the table look like in predictor_data.shp? There should be a column for each predictor variable with the observed values of that predictor in each polygon.

If a predictor has no data across all the polygons then it is dropped from the regression model. That would happen if there is no spatial overlap between the predictor and the AOI. Sometimes incorrectly defined coordinate systems can be the root cause there.

Hi Dave, thanks for the response!

The predictor_data.shp has zeros for the predictor variables that are not being included in the model. It is possible that I have incorrectly defined my coordinate systems. I currently have them set for WGS84 UTM zone 11 for Idaho and all predictors are matching the AOI. Do you have a suggestion for a different coordinate system? Thanks for the help!

Sarah

Well zeros should be valid data and not excluded from the regression, so that may be a bug. But do you think it’s correct for all of the predictor values to be zero in your case? If not I would double-check in GIS that all the layers are actually overlapping. If they are not, then it’s possible they have been incorrectly assigned that UTM system. The UTM system is a fine choice, but only if the coordinates are actually units of that system, and not say, decimal degrees.

It is hard to say if zeros are correct. I would think that it is incorrect, but it is possible that the Flickr data isn’t present in those areas. My GIS layers are overlapping when pulled into Arc so that doesn’t seem to be the problem but I have to admit that you probably understand coordinate systems better than I do.

Thanks for bearing with my suggestions. It probably makes the most sense to share all your input data at this point so I can reproduce the problem. Could you share a google drive, dropbox, etc, link with me? You may do so here or in a private message.

I really appreciate your help and hopefully it doesn’t take up too much of your time! Here is a link to a google drive folder: https://drive.google.com/drive/folders/1gEUDyknatNg3Z-w5mHgxbRxzoJFM2Ag-?usp=sharing

@sarahhalperin, thanks for sharing your data. It appears this problem comes down to a bug in the model and a table-formatting issue. There are some trailing spaces in the predictor table csv rows for the predictors that were being ignored. That caused the model to be unable to identify the polygon_percent_coverage keyword in the table. I removed those spaces and that solved the problem. Here’s the edited table:

PredictorTable.csv (300 Bytes)

We’ll fix that bug so extra spaces won’t be a problem in the future.

Thank you so much! That makes a lot of sense

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.