What is the global significance of a 5-sigma Higgs signal?

[mfb]

You can plot a curve "likelihood versus signal strength" - that will give (approximately) a gaussian curve with the maximum at the observed signal strength and "zero signal" about 5 standard deviations away. The real analysis is a bit more complicated, but that is the basic idea.

 

[Vanadium 50]

Vixra is not an acceptable source here, so please be careful with it. I am letting it in here because it's being used as an example for a certain kind of plot, but normally it cannot be used.

 

[exponent137]

You can plot a curve "likelihood versus signal strength" - that will give (approximately) a gaussian curve with the maximum at the observed signal strength and "zero signal" about 5 standard deviations away. The real analysis is a bit more complicated, but that is the basic idea.

Yes, let us say simplified, that is that Higgs signal gives some gaussian curve with center close to 125 GeV and with width somewhere <1 GeV, thus 5 sigma < 5 GeV.
But, what is zero signal? Is it some grass everywhere on energy scale and this grass has some bumps?

I am not sure if we should look 5 sigma of those bumps, but Higgs signal should be out of those 5 sigma?

 

[exponent137]

You can plot a curve "likelihood versus signal strength" - that will give (approximately) a gaussian curve with the maximum at the observed signal strength and "zero signal" about 5 standard deviations away. The real analysis is a bit more complicated, but that is the basic idea.

On blog, Gibbs writes many reports from Higgs measurement and he does not write many speculations. viXra archive is not the same as his blog. Similar surnames similar topics :) .

 

[mfb]

Yes, let us say simplified, that is that Higgs signal gives some gaussian curve with center close to 125 GeV and with width somewhere <1 GeV, thus 5 sigma < 5 GeV.

No, that is a completely different thing.
The significance is related to the signal strength ("number of Higgs events in the peak") and its uncertainty. Assuming gaussian errors, this uncertainty would be about 20%. This is called "local significance", as it is the significance of the peak at that position.

Zero signal would be "background-only model fits perfectly". Of course, that will not happen due to statistical fluctuations, so you expect to see some 1-2 sigma deviations in many different spots. However, you do not expect to see a 5-sigma deviation. If you calculate the (small) probability to get a 5-sigma signal by random fluctuation, you can calculate a "global significance" of your 5-sigma result.


@post 39: I think you quoted the wrong post there