small vignette for sample handling order randomization

[julianesiebourg]

The vignette uses a X2 test scoring function for sample handling order randomization

[julianesiebourg]

@ingitwetrust, @idavydov and @banfai,
this is a first try to deal with sequence blocking.
I think there's room for improvement as the solution in this toy example is not perfect.

[idavydov]

Hey @julianesiebourg , apologies for such a long silence. I made a few cosmetic changes.

The part which is not very clear is the score. I think I understand it, but it would be amazing if you could add a few more explanations there.

[idavydov]

Another suggestion: we do not have ComplexHeatmap in Suggests; this breaks building vignette.

Since the plot is quite simple, do you mind redoing it with e.g. geom_tile()/geom_raster()?

[julianesiebourg]

Hi @idavydov,
I removed unnecessary dependencies and added some description to the score as well as an evaluation and why we are not super happy with this solution.
This should serve as a starting ground for finding better solutions.

[idavydov]

Hi @julianesiebourg ,

I added another approach (based on neighboring samples). Could you please have a look?

[julianesiebourg]

Hi @julianesiebourg ,

I added another approach (based on neighboring samples). Could you please have a look?

Hi @idavydov,
This looks indeed very nice (actually, close to perfect ;) ).
Should we provide this as a built in scoring function for the users? Similar to osat? I think it could be reused a lot. What do you think?

On a bit of a different note. I'm still wondering if we could handle a 2D layout in a similar way. Is this the same as what we tried for plate scorings (there we endet up still getting clustered results). Or is this the current solution different but could be extended for 2D?
(I'm also currently looking into this score for 2D layouts https://en.wikipedia.org/wiki/Moran%27s_I but havn't gotten my head around how different this is, yet).

[idavydov]

@julianesiebourg I implemented mk_min_distance_score(), could you please take a look?

Somehow you're not a member of bedapub, so I cannot set you as a reviewer. I also just invited you.