META TOPICPARENT |
name="MDH0000" |
Useful Analysis Tips
Error bar estimation of Two efficiency multiply
y = f * g (f : Eff1, g : Eff2)
dy = (@y/@f)df + (@y/@g)dg
Relative error
dy/y = [(@y/@f)df + (@y/@g)dg]/fg
here, y = f*g, so (@y/@f) = g, (@y/@g) = f
dy/y = df/f + dg/g
|dy| = sqrt( (df/f)^2 + (dg/g)^2 )
final value is y + | dy | = y + sqrt( (df/f)^2 + (dg/g)^2 )*y
Therefore
the error bar of multiplied two variables
is sum of the each error of those.
Dividing error:
y = f/g >> dy = (f/g)' = f'/g - g'/g^2
dy/y = (f/g)'/(f/g) = f'/f -g'/g
| dy | = sqrt( (df/f)^2 + (dg/g)^2 )
final value is y + | dy | = y + sqrt( (df/f)^2 + (dg/g)^2 )*y
efficiency of pair
The factor 2 comes from the from the fact that the two single muon efficiencies that make the pair efficiency are fully correlated.
In that case the uncertainty on pair is just the sum of the uncertainties on the single efficiency. You can also simply derive it like this:
for eff_pair = eff_single^2
-> err_pair = sqrt( (deff_pair/deff_single)^2 err_single^2 )
= sqrt( (2*eff_single)^2*err_single^2)
= 2*eff_single*err_single
that's the expression for the absolute uncertainty, to get the relative uncertainty, we just divide by eff_pair
-> err_pair/eff_pair = 2*eff_single*err_single/(eff_single^2)
= 2*err_single/eff_single
sqrt((eA/A)^2 + (eA/A)^2 + 2 (eA/A)^2) = sqrt( 4*(eA/A)^2 ) = 2 * eA/A
|