Statistics code for desktop (and soon mobile)

[Willem Ferguson]

> On 09.01.21 00:13, Berthold Stoeger wrote:
>> Hi Christof,
>>
>> Thank you for the detailed report. A few short comments in-line.
>>
>> On Freitag, 8. Jänner 2021 23:25:07 CET Christof Arnosti via subsurface wrote:
>>>    * When I select date(yearly) as base variable and buddies as data, bar
>>>      charts have a yellow warning in the drop down. Why's that?
>> A bar-chart is not recommended with continuous data. A histogram is preferred.
>> However, as you note, the warning icon is not a good UI element.

I suspect the warning bar is indeed a good UI element. The difference 
between continuous and categorical variables will be defined in the user 
manual, once I get to writing that text. It is not good practice to mix 
the use of these two different types of information and a solid warning 
is the least that I would expect. Personally I would prefer a red cross 
over each of the icons for the inappropriate graphs.

>>
>> In case of sparse data I have likewise found bar charts to be more convenient
>> than histograms. On the other hand, I see the argument that they can be
>> misleading if only a few cups of data are missing.

Zero data are useful information. If you did no dives between 30 and 50 
meters, but yet did do dives down to 60m, then the blank areas in the 
range 30 - 50 meters represent valid data: no dives were executed for 
those depths. To ignore zero data is to misrepresent the information. 
This is one of the cases where using a bar graph to represent depth data 
is statistically wrong, leading to erroneous conclusions.

>>
>>>    * The trend line does not always appear in the scatter graph. For
>>>      example, when I select date (no binning) / depth, there is no trend
>>>      line, except for when I filter out very shallow depths. For water
>>>      temp over date I get a trend line right away. I'm sure that's
>>>      correct and there is a statistical reason for this that I'm not
>>>      aware of.
>> Indeed, there is a statistical test whether there is a linear regression.
>> Willem knows more.
>>
The red trend line (=linear regression line) is only shown for 
scatterplots where there is a statistically significant relationship 
between the two variables that are plotted. When there is no line, there 
is no relationship. Presenting a trend line in this case would be 
misleading and leads to erroneous conclusions. This will be explained in 
the user manual.

>>>    * When I select Buddies over Date(yearly), and then grouped vertical
>>>      bar chart, the bars are oddly spaced. I suspect that for every buddy
>>>      there is a bar every year, even if the number is zero. This might
>>>      make sense in some cases (for example water temperature), but in the
>>>      buddy case it looks weird. Maybe add some "don't show empty bars"
>>>      option for the grouped bar charts?
>> I'm not an expert in charts, but I think that this is how grouped bar charts
>> should be done(?). In the case of sparse data a stacked chart is probably
>> preferred.
In this case the discussion above is pertinent.*_Zero data are real 
data_**.* *Data of absence does NOT mean absence of data.* If a bar is 
empty it indicated that the buddy did not dive that year. The 
information indicating you did not dive with buddy X during 2020 is just 
as valuable as the information that you dived with her 30 times in 2019. 
Why make the graph pretty by eliminating the bars with zero height??
>>>    * Is there some Export functionality planned? For example simple image
>>>      export of the graph?
Screen capture software can easily be used to save a graph in almost any 
graphical format except as vector graphics.
>> Not yet.
>>
>>>    * For me the Filter GUI seems a bit unintuitive. When there is no
>>>      constraint present, it's not very obvious that constraints can be
>>>      added (the button is in an odd place). A change to make it more
>>>      obvious could be to add a "Constraints" heading below the fulltext
>>>      search, and move the button there? And maybe also display a "No
>>>      constraints" text when no constraint is set? I really like the
>>>      "Filter sets" functionality!
>> Yes, I also noticed that - especially in the German translation - the filter
>> is quite inaccessible.
If you click on "Add constraint", a list of all possible constraints is 
shown. I Need to look at the user manual to see if this aspect has been 
explained clearly. This is possibly my fault.
>>
It is fantastic that you actually looked at the statistics option and 
that you  spent time to think about how it works or how it could work. 
Your comments are extremely valuable. Once the user manual for this has 
been written, hopefully many thing will become more clear, but your 
interpretation of the information and defining your needs are very 
valuable. Thank you for doing that.

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.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.subsurface-divelog.org/pipermail/subsurface/attachments/20210109/4f7594ca/attachment.htm>