Activity Reports Correlation

[admin]

Hi OMB.

A question about your correlation charts. Are they implemented in activity reports for multiple workouts, or only in activity details ?

What I would like is a x/y scatter plot where I can select a field for each axis from the activity. For example, on the Y axis show the average HR, on the x axis show the average pace. The axis values might also be numeric custom fields.

Or... do you know another plugin which does this?

[old_man_biking]

Hi OMB.

A question about your correlation charts. Are they implemented in activity reports for multiple workouts, or only in activity details ?

What I would like is a x/y scatter plot where I can select a field for each axis from the activity. For example, on the Y axis show the average HR, on the x axis show the average pace. The axis values might also be numeric custom fields.

Or... do you know another plugin which does this?

I have implemented both
* my Training Analysis plugin provides the correlation WITHIN ONE activity. See http://www.sporttracks-plugins.com/Spor ... ocId411250
* my Activity Documentation and Reporting plugin provides the correlation ACROSS ACTIVITIES OVER TIME. See http://www.sporttracks-plugins.com/Spor ... ocId601856

Cheers,
OMB

[admin]

It's very close. My particular scenario requires custom ATHLETE fields (both number and timespan), which seem to not be supported. Other than that, it's what I needed.

[old_man_biking]

There is one issue, when you try to put athlete history specific entries into an activity report:

An activity report has one line per activity (+ rollups).
When you add athlete info specific entries as columns,
and there is more than one activity for a day,
then there'll be multiple entries for the same athlete history entry in the report.

That doesn't hurt much, when the aggregation type is "average".
But when the aggregation type should be "sum" (e.g. for something like calory intake,
the report will have incorrect rollups (4 activities a day let the calory intake multiply by four.

I don't know what you intend, but maybe this little treatment helps

Cheers,
OMB

[stavros]

For example, on the Y axis show the average HR, on the x axis show the average pace. The axis values might also be numeric custom fields.

Other than the custom Athlete data, it sounds like you are trying to achieve the same thing I wanted when I requested the enhancement...
http://www.zonefivesoftware.com/sporttracks/forums/viewtopic.php?f=3&t=7214

The original solution from OMB in ST2 delivered exactly what I was looking for, and then improved further in ST3 with the additional filtering available through Activity Reports.

Here's an example of a plot I would look at regularly. It's the same x and y axis you suggest, filtered to a standard course I run, and filtered to activities with 'clean' data tracks (I auto-populate an equipment item called 'good data' for new data downloads and can delete if there was a problem with data like heart rate spikes). From the chart I can validate (or not) my perceived fitness without having to do a full on time trial. For example, the latest run highlighted had a slow 'active' average pace, but the associated 'active' average heart rate was proportionally low. You can see in comparison the run before last (9-Jan) was on the 'wrong' side of the diagonal trend line, which provides some evidence that a change to my training approach over the last six weeks has helped improve my fitness / performance and move me to the 'right' side of the trend line.



Is this what you are looking for?

[admin]

This is what I ended up with. I couldn't use custom athlete fields, but I got close enough.

I personally found this chart to be very confusing. This is no knock on OMB - the plugin does exactly what it advertises. However as a trend spotting tool, it requires a lot of brain processing - you must understand the axis, the trend line and the color "movement" simultaneously. I think in part this is caused by two factors: a) There are too few data points, b) The data "moves" too quickly over time (colors are not clustered).

Looking at your chart I just had an idea. Take the same plot and measure the distance (+ or -) from the average line, then plot that +/- on a vertical axis with the time across the bottom (more recent to the right). You'd see a familiar time trend, with negative numbers indicating improvement over the average.

[stavros]

Your made me laugh! I had contemplated exactly the same idea about measuring the distance from the trend line but I couldn't be bothered digging out my fourth year engineering mathematics text book to figured out the formula to put in the 'Calculated Fields' plugin to give me that solution. So it would seem I'm not the only data geek. Anyone else willing and able to solve the problem?

[old_man_biking]

...
Looking at your chart I just had an idea. Take the same plot and measure the distance (+ or -) from the average line, then plot that +/- on a vertical axis with the time across the bottom (more recent to the right). You'd see a familiar time trend, with negative numbers indicating improvement over the average.

A fine suggestion. If I understand right, "from the average line" means "from the correlation line". Will give it a try.
Funny, some of my "more sophisticated" users have even suggested to add another dimension to this chart, making it even more complex

[admin]

Sure, you have bubble size and shape too, 6 dimensions should be easy

[old_man_biking]

Yes, and with these mouses that give user feedback by vibration, I'll create informative feedback on mouse over:
- power < 100W: gentle vibration
- power < 200W: this rocks
- power > 500W: don't ask