lgbouma / cdips

Tools for the Cluster Difference Imaging Photometric Survey

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Update PCA routine

lgbouma opened this issue · comments

The processing breaks at line 143 of cdips/lcproc/reformat_lcs_for_mast.py where there's a break in the code with:

    # TODO: replace this entire PCA routine with a call to                      
    # dtr.calculate_linear_model_mag, as in tests.test_pca. The routine below  
    # was good for S1-S13; we will update it in upcoming reductions

For the Cycle 2 reductions, it would make sense to:

  • use fewer principal components (like 5 max) than previously -- better for rotation period extraction.
  • add an ability to use other vectors to decorrelate against, e.g., background timeseries, CCD temperature timeseries etc. The background timeseries was the only one that wound up being useful.
  • add time-smoothing to the PCA vectors

This is mostly already done in cdips/lcproc/detrend.calculate_linear_model_mag, just needs to be merged.

Resolved by commits c4b0023 through 7e27cfc

Example randomly selected light curve from S14:

temp