Minutes of the E614 Software Meeting

                          7 February 2000


 Discussion of blind data analysis :

       As a means of removing bias from the analysis of the final data 
       sets a blind analysis of simulated data has been proposed.  Various 
       physical parameters (such as Michel parameters) would be slightly 
       altered then used to generate Monte Carlo data.  Analysis of this 
       data would proceed without prior knowledge of the parameter values.
       Once the analysis produced the correct physical parameters, it 
       could be used on the real data.

       The same procedure might be used to check whether wire positions 
       are correctly extracted from the calibration data.  In this case 
       Monte Carlo data would be generated in which wires were displaced
       from their nominal positions but only the generator of the MC data 
       would know the displacement values.

       Background subtraction would also be checked using a blind analysis
       of Monte Carlo data.

       There should be a logical separation of the tracking routines from
       the response function fitting routines which will be used to 
       extract the Michel parameters. 

       Large numbers of simulated events (about 10% of the real data sample)
       will need to be generated in order to correctly reproduce detector 
       response functions.  It has been suggested that the Pentium farm at
       Alberta could help out in this regard.  

  
       Thus far it has been conlcuded that no change in the analysis code
       is required to accommodate a blind analysis as long as Monte Carlo
       history banks are included in the simulated data.

Created for the The Center for Subatomic Research E614 Project Projects Page.

Created by The CoCoBoard.