arc XWSIGMA drill down error

desihub / qlf

Developement of the QLF framework for DESI

BSD 3-Clause "New" or "Revised" License

4 stars 4 forks source link

arc XWSIGMA drill down error #265

Open sbailey opened 6 years ago

sbailey commented 6 years ago

Testing http://desi-1.kpno.noao.edu:9090 using code from desispec #702 (now tag 0.25.0) and qlf #253, mostly looking at Observing History window and drill-downs.

arc exp 3571 Step 1 XWSIGMA has a warning, but the drill down view has "Can't load QA: object of type 'NoneType' has no len()" so I can't tell what is wrong.

felipelm commented 6 years ago

@sbailey we fixed this issue please verify at http://desi-1.kpno.noao.edu:9090/

sbailey commented 6 years ago

I no longer see the error message, but I'm not sure the values in the plots are meaningful. Please check if this is due to QL outputs, or due to the interpretation of those outputs. If the problems are in the QL outputs themselves, please open a ticket in desispec and then go ahead and close this:

X std dev and W std dev have some fibers that are hundreds or thousands of pixels?
Middle histograms could use some tuning so that the huge outliers don't dominate the bin size choices
XSIGMA AMP and WSIGMA AMP have 'nan' for amps 1 and 2 (I don't think per-amp values for these metrics are meaningful anyway)

felipeaoli commented 6 years ago

Hi Stephen, Thanks for your review. Regarding your points: 1) X std dev and W std dev have some fibers that are hundreds or thousands of pixels? We are receiving this values from quicklook.

**2)** Middle histograms could use some tuning so that the huge outliers don't dominate the bin size choices

I tried here some options, but the outliers are extremely larger than median. So, I coded a modified Z-score method function but in this case it seems better to show the outliers in a different histogram (in the same plot).

**3)** XSIGMA AMP and WSIGMA AMP have 'nan' for amps 1 and 2 (I don't think per-amp values for these metrics are meaningful anyway)

I kept the plot because the is still in the ql results for this flavor, but we can remove it. Regarding the nan's it seems something strange in the shape of list XSIGMA_AMP (e.g. [[nan, nan, 0.242, 0.240], [nan, nan, 0.247, 0.248]] )

@Srheft , could you please help us with the items 1) and 3) ?

image

Srheft commented 6 years ago

Hi @felipeaoli could you use PR#707 and see if you still see the NaNs in XWSIGMA_AMP? let me know for what exposure it occurred. Thank you.

Srheft commented 6 years ago

@felipeaoli if this PR behind the master, once the changes in the master are in this PR, those spikes in the XWSigma histogram should be removed. Or if you run master on an arc exposure, you should not see the spikes; please let us know if you still see them then. Thanks.