Hi guys, Owen, see if you agree with the following strategy. I think we need to make a test of the end-game and set up the machinery to do so, and that this should be the highest priority for Harold in addition to helping Sudan get the tau->hadrons analysis working.

We need 5 MET templates T_i:

1) T_1 = all-leptonic ttbar

2) T_2 = semileptonic ttbar->e,mu

3) T_3 = semileptonic ttbar->tau-> hadrons

4) T_4 = semileptonic ttbar->tau->e.mu

5) T_5 + all-hadronic ttbar

Step A

Define an overall template T that we can compare (normalize) to data.

T = Sum_{i=1,5} f_i * T_i

where f_i is the relative contribution of T_i to the N-1 MET plot in our final sample as determined from MC. We should already have these numbers. (They will change a bit if we add a tau DLV, but we can ignore this for now.)

Step B

Normalize the overall template T to the data using the MET region 60 < MET < 100.

Step C

Go through the numerical exercise outlined by Owen in his letter of Sept. 1 to determine our estimate of the number of ttbar background events for MET>150 and compare this to the true number determined from cheat info.

General

We currently have a control sample only for template T_2. For the time being, we can use the cheat info with the standard sample for the other 4 templates. We can replace the cheat info with templates as we develop them.

We can divide the ttbar samples into two, using half to define the templates and the other half to define the "data." To test the mechanics and verify that the closure test is working, ignore susy contamination for the time being (we will add this later). Thus use pure ttbar event samples.

Any comments, criticisms or better ideas ?

Owen (Sept 7th):

=================================================

- For step A, do we want to go all the way to zero in MET? We will eventually need to include non-ttbar BGs as well. Would it be better to instead define the fractions (f_i) in the normalization interval of step B?

- Are there any reasons not to extend the normalization interval from 60 < MET < 100 to 60 < MET < 150 ? We will probably be statistics limited, so making the window as wide as possible might help.

- In the exercise, we will need to include some uncertainties on the relative fractions (f_i). In order to get a realistic final uncertainty, we need to think through how we will get these and how our choice for the normalization interval in step B impacts these fraction uncertainties.

=================================================

-- HaroldNguyen - 08-Sep-2010