Dodgy signals about selection-atlas

[alimanfoo]

I think there are at least two issues to tease apart here.

One is that the threshold we use for reporting signals (delta_i) may be set too low, and so we are reporting signals that we feel are two weak or not well supported. I think that is definitely the case, and so some recalibration of this threshold seems in order.

A second is that I am seeing a small number of cases where I can't quite see how any kind of exponential peak was fitted to the data, because there is no visible peak shape there, even a weak one. I am more concerned about investigating this case first, because it may suggest other features of the peak fitting algorithm need adjusting, such as the amount of skew we allow.

For the moment, I would very much appreciate you posting any examples here of this second type of problem, i.e., examples where you can't see any visible evidence for an exponential shape peak.

[alimanfoo]

Here's another one...

[alimanfoo]

I'll have a look into this one...

[sanjaynagi]

There seems to be quite a lot of these @alimanfoo ....

Any idea why the signal detection algorithm would have got worse? ....

[sanjaynagi]

60-85 Mb

---

Uncalled signal on the left looks more real than called one on the right:

---

I've seen a couple like this...?