Raster Statistics

timstoebner · ‎05-03-2014

I have several rasters that are 32bit float pixel type. The statistics under properties in Catalog are min=0, max=1, mean=-1.
Obviously this cannot be. When I classify the rasters under "symbology" in Arcmap the statistics are reported correctly for example min=0 max=1 mean=.58 and pixel values appear to be accurate. I am trying to run correlations between the rasters using the "Band Collection Statistics" tool, but does not seem to work (I assume due to the incorrect values under properties). I tried using "calculate statistics" in catalog, but that did not help. I tried clean statistics and rerunning calculate statistics, but no difference. Any ideas?

ArcInfo 10.0 license.

Tim

NeilAyres · ‎05-05-2014

What format are these rasters?
Are you saying that after using the calc statistics tool, the values shown under properties are still as they were?
It is possible that the directory or file setting are set to "read only". Modify this using the normal file explorer.
I have seen in the past that calc statistics can run and complete normally, but fail to update the metadata files.
Good luck,
Neil

timstoebner · ‎05-18-2014

What format are these rasters?
Are you saying that after using the calc statistics tool, the values shown under properties are still as they were?
It is possible that the directory or file setting are set to "read only". Modify this using the normal file explorer.
I have seen in the past that calc statistics can run and complete normally, but fail to update the metadata files.
Good luck,
Neil

Neil,
The rasters are tif files. Even after exporting to GRID the stats do not change.

Yes, after running calc stats the problem persists. I even tried clearing the stats in catalog and running calc stats again, but no change.

I don't think it's a read only problem, because when I converted the original probability values to percents ("raster" times 100) the problem persisted in the subsequent output rasters. It was only after I converted the percent rasters to integer that the stats were finally normal.

You should also know that I have seven sets of these rasters that I calculated in model builder using map algebra and this is the only set (every raster in the set) that is exhibiting this problem. I have double checked the map algebra and re-ran several new rasters using a new model, but got the same results. Can it be caused by one of the input rasters in my model?