labels for trend lines are not clear

[GoogleCodeExporter]

The linear trend lines generated (a dashed line for the entire time range, and 
a solid line for ? last 30 years?) could use better labeling.  Instead of the 
current label:
   Trend (degreesC/Century) and R^2
   full: 0.57 (1.61)

How about:
   Linear trend lines as degreesC/Century (goodness of fit, R^2)
   Full time interval: x.xx  (y.yy) ; Last 30 years:   z.zz  (w.ww)

While looking into this issue, I noticed a comment that there is a bug in the 
trend line code when there is only one data series.  I looked into this and 
think I have found a bug.   Here is a diff:

tool$ diff vischeck.py orig.vischeck.py 
139,140c139,140
<     slopes = [t[1:3] for t in trends]
<     coefficients = [t[3:] for t in trends]

---
>     slopes = [t[1:len(data)+1] for t in trends]
>     coefficients = [t[len(data)+1:] for t in trends]

This fix seems to work, but when testing the --offset option with multiple 
copies of the same data series, the trend lines seem to still overlay one 
another.   I'm not sure if this is another issue, or if my change has impacted 
that feature.   Looking at the code, I don't immediately see how my change 
would impact that, but I must admit I don't understand the use of "scale" in 
this code. Trying 3 copies of the data series blows the 'get' character limit 
that the Google chart docs warn about ... perhaps one of you developers can 
carry this forward?    Thanks, Terry

Original issue reported on code.google.com by terryhei...@gmail.com on 2 Aug 2010 at 11:02

[GoogleCodeExporter]

Thanks for the bug report.  Er, 2, really.

I'll think about the text clarification that you suggest.  My concern is using 
up lots of space on the chart, but one nice aspect of recent warming is there 
seems to be plenty of space on the bottom right of most charts.

These aren't intended to be journal quality graphics (indeed, more as a "visual 
check" that the results are plausible), so I'm not inclined to spend a huge 
amount of effort on it.

Your second bug: yes, there is a bug with only one series.  I will get round to 
fixing it at some point, your diff will be a very useful starting point, thanks!

Do you actually use the --offset option?  The code is descending into an 
unusable pile and I'm inclined to remove features to make it simpler.

Original comment by d...@ravenbrook.com on 18 Aug 2010 at 9:58

[GoogleCodeExporter]

Thanks for the reply.   My original report did not make clear what I really 
found confusing about the current labeling:  the "misdirection" caused by the 
different location of the parentheses in the caption:
       Trend (oC/Century) and R^2
and in the data lines:
       full: x.xx (y.yy) / 30-year: z.zz (w.ww)

In the caption, the parens surround the definition of the units of the trend, 
but in the data lines, the parens surround the R^2 value!   

So, to be less wordy, how about this:
       Trend in oC/Century (R^2)

I agree it is picky, but aren't we trying to be _clear_ ?

> Do you actually use the --offset option? 

No, I only used in while testing the suggested bugfix.  I was lazy and wanted 
to reuse the current series, not find another series ;-)

Original comment by terryhei...@gmail.com on 19 Aug 2010 at 11:54

[GoogleCodeExporter]

I agree the use of parens is unclear, I'll change it.

I agree your diff is a good fix.  The code was terribly unclear and I've tried 
to improve it.  That meant I didn't use your diff, but it was very useful.

I added Issue 87 for the missing slope display for single series (and will 
shortly fix it).

Original comment by d...@ravenbrook.com on 5 Sep 2010 at 8:49

[GoogleCodeExporter]

This issue was closed by revision r553.

Original comment by d...@pobox.com on 5 Sep 2010 at 9:00

• Changed state: Fixed

[GoogleCodeExporter]

Terry, as per your suggestion (more or less), I've changed the display title to 
"Trend in °C/century (R²)" and the trend values to "full: 1.06 (0.72); 
30-year: 1.57 (0.66)"

If you still have the energy please can you comment on whether you think it is 
clearer?

Original comment by d...@ravenbrook.com on 5 Sep 2010 at 9:04

[GoogleCodeExporter]

Yes, I think it is clearer now.   Thanks.    

The "30-year" doesn't really convey the "last 30 years", but with the visual 
help of the presumably shorter 30-year-long slope on the graph, it is probably 
ok, but...   

if you can spare a few more chars in the display title, I still think "last 30 
yrs" would be clearer.

Original comment by terryhei...@gmail.com on 6 Sep 2010 at 1:16