1 quaero bruce knuteson berkeley/chicago an automatic model-tester a new way to publish hep data

19
1 Quaero Bruce Knuteson Berkeley/ Chicago An automatic model-tester A new way to publish HEP data

Upload: sydney-boyd

Post on 20-Jan-2016

221 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

1

Quaero

Bruce KnutesonBerkeley/Chicago

An automatic model-tester

A new way to publish HEP data

Page 2: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

2

Motivation

How do new physics searches usually work in this field?• Theorist writes down a model• Six months later an experimentalist decides it is worth checking• Two years later her graduate student finishes the analysis• Six months after that the publication comes out

Is it possible to this in 3 hours, rather than 3 years?

An automatic model-tester

Page 3: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

3

Where do we spend most time during an analysis?1. Understanding backgrounds to data2. Generating signal Monte Carlo and associated bookkeeping3. Optimizing cuts and setting limits

Not a whole lot we can do about #1 . . .

But can we offload #2, and automate #3?

MotivationAn automatic model-tester

Page 4: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

4

Page 5: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

5

Quaero final statesAn automatic model-tester

Page 6: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

6

The details of the Quaero algorithm have been presented in New Phenomena and Run I meetings

The basic idea is quite simple:

The algorithmAn automatic model-tester

1. Construct a background estimate

2. Construct a signal estimate

4. Choose Dcut to optimize an expected 95% limit

3. Define D=p(s)/[p(s)+p(b)]

Page 7: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

7

But does it work?

ExamplesAn automatic model-tester

Example analyses performed:

WW eET

ZZ eejj

h WW eETjj(nj)

h ZZ eejj

tt eEtjj

tt eETjjjj

Z’ tt eETjjjj

LQLQ eejjDetails available in draft Quaero PRL

Results all appear reasonable, with no fine-tuning

(w/ Greg Landsberg)

Page 8: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

8

An automatic model-tester

A new way to publish HEP data

Page 9: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

9

Now that Run I is winding down, how can we preserve our data in an accessible form?

• Our data are “context-specific”(You need to understand an awful lot about it

in order to do anything with it)

Is there a way that we can make our data easily available, both to ourselves and people outside

DØ?

MotivationA new way to publish HEP data

Page 10: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

10

Astrophysicists appear to have solved this problem

High energy physicists so far have not, despite serious attempts (e.g. at LEP)

Perhaps DØ has hit upon a solution?

MotivationA new way to publish HEP data

Page 11: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

11

Page 12: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

12

Advantages include:# of models outnumbers # of us by lots

Put theorists to work

Perhaps we missed something?Never hurts to have another pair of eyes

Increase number of DØ citationsReferences to the papers describing the data

Education / outreach via QuarkNetSome additional work (niftier interface), but great PR

potential

Frees us up to focus on Run IIBut lets us make the most of Run I

Facilitates communication with theoristsShould help bridge the theory/experimental gap

DØ would be leading HEP in making its data accessible

Other experiments have tried (and failed) to solve this problem

Allows us to beat any CDF analysis by at least a year

Quaero has the answer in a matter of hours.

PossibilitiesA new way to publish HEP data

Page 13: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

13

There are a number of ways Quaero could be implemented

PolicyA new way to publish HEP data

“Put the data out there”Make data available to all with no internal review

Make data available with limited scope and internal review

Restrict those who are allowed to use QuaeroReview all Quaero results before releasing them

Don’tKeep Quaero as an internal tool

Make data available with general scope and more limited internal review

many variations on these themes

Page 14: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

14

Representative comments/concerns from collaborators(Opinions were uniformly thoughtful and reasoned —

thanks to many for valuable discussions)

DØ worked incredibly hard for Run I data. Do we really want to “give it away”?

What responsibility would DØ be shouldering? What if some nut claims an unfounded discovery?

Does this set a dangerous precedent for Run II? Would this have a detrimental effect somehow on exp

HEP?Does DØ have the resources to conduct an internal review of all Quaero results?

Who would do it? What would the rules be?Does making Quaero results available only after internal review seem slimy and unforthcoming?

PolicyA new way to publish HEP data

Page 15: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

15

My view/proposal:

We have a real opportunity here

There are legitimate concerns about making DØ data publicThis has not been done before

These concerns can be addressed (I think)

I began this effort believing that an internal review of Quaero results was necessary

Does DØ have resources to commit to this effort? No.Would it be cleaner to provide results directly? Yes.What if there are bugs in Quaero? Make sure there aren’t.

I currently feel that the cleanest approach (with greatest potential advantages) is to “put the data out there”

PolicyA new way to publish HEP data

Page 16: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

16

The present plan:

EB 139 is reviewing the accuracy of the Quaero method and example analysesQuaero was made available for general New Phenomena group testing beginning December 2000 All are welcome to test Quaero beginning today.http://www-d0.fnal.gov/~knuteson/d0_private/quaero/(From Feb 15th onward Quaero will also generate signal for you.)

Comments and suggestions are actively solicited

PRL draft existsDescribes Quaero method, data, and results of examples

Hope to simultaneously publish in PRL and release Quaero to the HEP community

PolicyA new way to publish HEP data

Page 17: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

17

Quaero

An automatic model-tester

A new way to publish HEP data?

Conclusions

Page 18: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

18

Page 19: 1 Quaero Bruce Knuteson Berkeley/Chicago An automatic model-tester A new way to publish HEP data

19