Analysis
Analysi of ITk data requires special attention.
Some \(\chi^2\) fit based appraoches are either not optimal or even not applicable for the raw scan data used in ITk. In case of a correct model with Gaussian distribution of errors one would expect to get \(\chi^2/\)n.d.f. \(\approx 1\). When it is not the case, two options are possible:
-
If \(\chi^2/\)n.d.f. \(\gg 1\) this means, that the model does not describe data.
-
But, if \(\chi^2/\)n.d.f. \(\ll 1\), this does not mean, that the model perfectly describes data, this just means, that either uncertainties are overestimated, or the model is not applicable at all.
The second is the case, when one uses \(\chi^2\) fit for S-cirve functions.
Related links:
Maximum likelihood and significance tests [pdf]
Parameter estimation chi2 and likelihood [link] [pdf]