diff --git a/posts/template_with_distortion.ipynb b/posts/template_with_distortion.ipynb index dd4bf1f..d9381dd 100644 --- a/posts/template_with_distortion.ipynb +++ b/posts/template_with_distortion.ipynb @@ -594,7 +594,7 @@ "cell_type": "markdown", "metadata": {}, "source": [ - "HESSE reports succeess, but the uncertainties on offset and scale are suspiciously small. We should not trust this result. This is one of the possible outcomes if the second derivatives are not well-defined bad. In other cases, HESSE may fail or report that the covariance matrix had to be forced to be positive definite." + "The minimum is reported as invalid, that is suspicious. HESSE reports success, but the uncertainties on offset and scale are suspiciously small. We should not trust this result. This is one of the possible outcomes if the second derivatives are not well-defined bad. In other cases, HESSE may fail or report that the covariance matrix had to be forced to be positive definite." ] }, { @@ -746,17 +746,13 @@ "At least in some cases, it is possible to fit distorted templates and estimate the distortion parameters as nuisance parameters, and get good results with modest effort. The solution described here will work best if the components in the mixture are well separated and if the same distortion can be applied to all templates (for example, because it models a calibration mismatch).\n", "\n", "In the following cases, this method will introduce large uncertainties on the fitted yields:\n", + "\n", "- the components are strongly overlapping\n", "- different distortions are needed for each template\n", "- the distortion function is very flexible and has many parameters\n", "\n", "In some of these difficult cases, reasonable results may be attainable if the distortion parameters are further constrained with penalty terms, that prevent the distortion function from diverging too much from the identity." ] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [] } ], "metadata": {