RFC: Statistics in Subsurface

[Willem Ferguson]

On 2020/05/17 21:51, Linus Torvalds wrote:
> On Sun, May 17, 2020 at 3:34 AM Willem Ferguson via subsurface
> <subsurface at subsurface-divelog.org> wrote:
>> As a side note it is in principle possible for case (5) and (6) above
>> to include a *continuous* 3rd variable e.g. a scatter plot of SAC
>> versus dive depth where the colour of the dots indicate the water
>> temperature of each dive.
> The temperature vs SAC scatterplot would be one of the few statistics
> I'd like to see, because it's the one that I personally think might be
> the best correlation. Not that I've really ever tested it.
>
> You have those "SAC vs depth" and "SAC vs suit" correlations, but I
> suspect at least some of them end up being about temperature, where
> the suit choice (and depth) end up being incidental rather than
> causal. Also, I happen to think thaty both of those choices are not
> good scatterplot axis choices, because I'd consider them fairly
> discrete (while temperature and SAC rate aren't).
>
> So if we do fancy scatter-plots, I'd rather see them done the other
> way around. Do the continuous variables as the axes, and then color by
> the discrete variables (so you can see suit type by color if you
> care).
>
> And yes, depth may _technically_ be continuous, of course, but I don't
> much think "depth" makes sense all that much sense treated as a
> continuous variable. Nobody cares about 16m vs 19m, I suspect. But you
> might care about "deepest dives" as thirds or quartiles, because
> there's certainly a difference between a shallow dive and a very deep
> one (but different people will have rather different notions of what
> constitutes a "deep dive").
>
> So if your max depth of all your selected dives is 176 ft (to pick a
> random number that happens to be my two deepest dives), maybe it makes
> sense to divide your dives into 0-60 ft, 60-120 ft, and 120+ feet and
> give them three different colors, rather than a continuous axis.
>
> Of course, this is only if you want to show three different variables.
> If you only do two, then by all means do the depth continuos.
>
>                   Linus

You make perfect sense. Thank you for this input. Classes of depth 
should be easy to accommodate with Dirk's Granularity concept.

The word Granularity may need to change because, even though it is 
entirely intuitive for me, it is perhaps not intuitive for the 
first-time user. However, I have no idea which word may be better.

Kind regards,

willem



-- 
This message and attachments are subject to a disclaimer.

Please refer to 
http://upnet.up.ac.za/services/it/documentation/docs/004167.pdf 
<http://upnet.up.ac.za/services/it/documentation/docs/004167.pdf> for
full 
details.