C. Cahillane

Posted are the results for the total frequency dependent calibration uncertainty budget for LHO.  
These results are for before the updated actuation models at LHO.  I still need to get the updated PUM and TST drivers, but this should be close enough for now.

The method used to construct this result is as follows:
1) MCMC fit calibration model parameters to the reference models from Nov 12, 2016.  
2) Record samples from the MCMC.
3) Get residuals = (Meas/MCMC Fit) for ALL measurements, not just reference model, and find systematic errors via a Gaussian Process (GP).
4) GP outputs a mean posterior function (systematic error) and a covariance matrix (uncertainty) relating all training data.
5) Record both the GP systematic error and covariance matrix.
6) Now, assemble the response function results from the MCMC and GP by taking a random sample from the MCMC results and generating a potential systematic error from our GP mean systematic error + covariance matrix.  

The final results are showed in Plot 1.  All 1000 instances are plotted in white, while the mean and 68% confidence intervals about the mean are plotted in red.