Signal injection test

1. May 20, 2013

Alkass

hi

I would like to ask, what is all about a "signal injection test" ? How is it done and what one expects to see ? I ve read about it in some Higgs study/plots CERN has published

thanks

Alex

2. May 20, 2013

Staff: Mentor

You can simulate what you expect without a Higgs, and then you can add a Higgs signal and see how the measurements change - is it possible to see the difference? As far as I know, "signal injection" is just that.
The same is possible for all other particle searches, of course.

3. May 20, 2013

Alkass

Ok, I get it - So what happens if you see the difference or not and how you quantify it ? Does it change the statistical significance of the null hypothesis in the case you can "see" a differnece? Or one should think about it, like a "biased" search, thus you should consider an H1 hypothesis in such a case instead of a null one ? At the end, why one would want to do such a test ?

thanks
-a

4. May 20, 2013

Staff: Mentor

If you do not see a difference between data and "simulation without signal", you can set an upper limit on the signal strength. If a simulated signal with some specific signal strength would lead to a significant deviation (usually: 10% or 5% probability that a downwards fluctuation gives fewer events), this signal strength is excluded at 90% (95%) confidence level.

That leads to those colorful plots like this one.

Consider a point at 150 GeV (just at the left edge of "LHC excluded"), for example, it is close to 1, and 1 is exactly the Standard Model signal strength. The point has the message "if there is a Standard Model Higgs at 150 GeV, we should have seen more events with a probability of 95%".

Actually, this is not true. Following my description, every point would have a probability of 5% to exclude everything, even if there is a Higgs and the data is not sufficient to see it yet. Therefore, the method is modified a bit to include the sensitivity of the search, but I think this is an irrelevant detail here.