Statistics code for desktop (and soon mobile)

[Dirk Hohndel]

> On Jan 10, 2021, at 2:06 PM, Rick Walsh <rickmwalsh at gmail.com> wrote:
> I have just had a short play with the statistics with the latest daily Windows build (subsurface-4.9.10-260-gbd26d8407f28.exe) and I must say this looks great.  It is ridiculously easy to create meaningful (and meaningless!) graphs of my dive data.  Thanks Berthold, Willem and Dirk for going to so much effort to devise and implement this.

Standard comment that the implementation credit goes almost entirely to Berthold. Willem and I have each done a small bit - I have done some more on infrastructure stuff triggered by Berthold's work.

> A few nitpicks/suggestions:
> - When I select max or mean depth as a variable, the binning options are 'none', 'in 5 m steps', 'in 10 m steps' and 'in 10 m steps'.  The duplicate entry should be removed

I just merged Berthold's PR that fixed that.

> - As noted in other emails, the scattergraph trendline / linear regression is frequently meaningless.  It would be good to add a check box to show or hide the line (similar to the checkboxes for labels and legend for the stacked bar charts).  I think this is probably simpler and more useful than determining base on the R value (or similar) whether the fit is good enough to make displaying the line worthwhile, but I have no objection to doing both.  Displaying the R value could also be appreciated by a few people, but others could find it annoying/confusing - if you do decide to do this, please make it optional.

This is currently under active debate... in the end I'll give a lot of weight to Willem's preference, but everyone's input is important as it helps shape our perception what users would actually care about.
I kinda like the regression lines when they show a clear regression... but I have seen a few cases where they seem rather questionable in their validity.

> - When hovering over a group in the bar or pie charts, it would be nice if the box that pops up lists the dives (especially when the list is short).  I have one dive with a SAC 68-70 l/min, and it'd be handy if the graph showed when it was.  There are also 7 CCR dives in my log, which is odd since I know I've never done a CCR dive.

I'd point out that it should be REALLY easy to identify these dives using the filter.
I'm not sure I really want to overload the info box this much. Curious what others think.

> - The variable of 'Buddies' seems to pick up both buddies and divemasters.  Apparently "Me" is one of my most common buddies (I'm not such an egomaniac that I list myself as my own buddy, but I did list myself as the DM).

This has been pointed out by others. Right now it says 'Buddies' but implements 'People'. It should offer all three choices.

> - An export / copy graph feature would be nice (but a low priority)

Both I and others have mentioned that it seems that every OS has a solid screenshot feature that really should give you all you need in almost all cases. So yeah, lower priority as far as I'm concerned as well.

> Other comments:
> -  This is a great way to identify bodgy data.  As well as the bogus CCR dives I mentioned about, it looks like I have one dive (under a pier) in the 75-80 m range, and two dives with a water temperature of -17.8 deg C.  I know these are incorrect, and tied to issues downloading from my old dive computer (addressed a few years ago by Jef, but it seems I neglected to tidy up all of the incorrect data)

Funny, I also identified two data correctness issues by looking at the statistics and going "what the HECK???" :-)

> - The date (year) vs No. dives graph is depressing.

Yes. 2 dives in the last 15 months :-(

Thank you so much for all the feedback and input. Really, really appreciated!

/D
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.subsurface-divelog.org/pipermail/subsurface/attachments/20210110/3c70c052/attachment.htm>