Poor model selection on histograms

[supazel]

Suggested configuration according to BIC is wrong
In the Histogram analysis tap, state configuration, the suggested configuration according to the BIC criteria is wrong.

To Reproduce
Steps to reproduce the behavior:

1. Go to 'Histogram analysis'
2. Click on 'Start analysis'
3. See suggested configuration

Desktop (please complete the following information):

• Windows 10
• Matlab Version 2017a
• MASH-FRET Version 1.1.2

Susann Zelger-Paulus
susann.paulus@uzh.ch
20190822_error_BIC.pptx

[mca-sh]

Hi supazel,
Thank you for the report.
What is it you observed that makes you think the suggested state configuration is wrong?
May be it is a misinterpretation of the BIC plot.
On the pptx you attached, the axis BIC=0 is located on the right and the minimum BIC is obtained for 1 state (the longest bar).

[mca-sh]

Hi supazel,
I understand that the suggested configuration is not satisfying, and that by eye, two Gaussians would seem to describe the histogram better.
However, it seems to be a limitation from the method used in MASH-FRET.
Could you send me the corresponding .mash file so I can verrify this?
Thank you again for the report.

[pflanzenhueter]

Dear all,
can we close this case?
Best wishes
Richard

[mca-sh]

Hi Richard,
I was thinking of leaving it open to anybody who is having the same concern
Mélodie