Data smoothing explained

Frequenly Asked Questions

Moderator: FAQ Moderators

Data smoothing explained

Postby CHnuschti » Mon May 18, 2009 10:44 am

Question:

How do I use / what are the effects of data smoothing?

Answer:

SportTracks allows to smooth data.

1) Preliminary remarks
  • If you apply smoothing, it can be reversed or modified anytime in the basic SportTracks.
  • Smoothing is an overall setting affecting all activities, regardless of category.
  • When SportTracks directly downloads a record from a unit, it writes an own file (*.fitlog) for each activity, containing the raw, detailed, unaltered data "as delivered". See storage FAQ for common locations of storage. Regardless if you apply smothing with Sporttracks or not, these file are left untouched.

2) Smoothing settings

Set up the smoothing at Setting=>Display/Analysis=>Data smoothing with mouse left click (Settings FAQ point 5)).
Smoothing can be set separately for various values. To apply smoothing, check the box with mouse left click, then set the smoothing value with the slider (mouse left click slider, then hold and move). A value is shown next the slider, this way you can reproduce settings. To deselect smoothing of a value, the box MUST be unchecked.

Image


3) Effects of smoothing

The more smoothing you apply, the more the data will be evened out. Data spikes (high as well as low) will be progressively reduced with increasing smoothing value. In technical terms, a given point/value X will be averaged with the adjacent points/values on the left and right ("central moving average"). The more smoothing, the more adjacent points/values are included to calculate the average, which will be the smoothed value of point X.
See example of heartrate, left without smoothing, right with smoothing set to "50".

Image Image


Smoothing has NO effect on time and distance, they represent the base for the application of smoothing and remain unaffected. Therefore it is to note that smoothing has NO effect on the calculation of the average speed.

Effect of smoothing is different depending from the type of the values.
a) Heartrate, cadence and power data is used to determine averages and maximum values. The effect of smoothing on minimum/maximum values can be considerably (they will be increased/decreased), while the effect on averages is generally much weaker.
b) Elevation data is used to determine total climb (ascent/descent), the calculation is to sum up the elevation differences. Smoothing can have a considerable effect on these sums, it generally will decrease them.

Smoothing has NO effect on manually entered values, see the "manual values" FAQ.

Averages and maximum values in your lists, summaries etc. will be displayed according to your chosen smoothing, they will become the "VALID" values.
Also the graphs and its values will be displayed according to your smoothing, they will be more continuous and better "readable".


4) Additional remarks

Grade is derivative from elevation, hence the elevation smoothing applies also for the grade.
Last edited by CHnuschti on Fri Sep 18, 2009 4:07 pm, edited 8 times in total.
CHnuschti
 
Posts: 267
Joined: Tue Feb 13, 2007 3:28 am
Location: Switzerland

Postby CHnuschti » Tue May 19, 2009 9:46 am

Question:

Recommendations for data smoothing?

Answer:

In principle you should choose your smoothing settings according to the reliability of the measurement of the values. Heartrate and cadence for example are very simple to measure accurately, very likely, you best take them "as delivered".
Elevation and power is more complicate to measure accurately, to some extent it depends from the devices used. Here applying smoothing is more likely to bring the values into a "real" range.

However, if you like better "readable", evened out graphs, then smoothing is your tool.

If you are interested in "real" values, best approach is to calibrate your smoothing settings with a credible, reliable, independent reference measurement. Comparing values from different "obscure" sources/software may give you some indications, but not necessarily clarify which value is the accurate one.
Unaltered data of devices/sources from related, reputed brands generally are the best comparison sources, for example:
- for elevation: (digital) maps with smallest scale possible
- for power&cadence (cycling): SRM measuring system
- for heartrate: Polar heart rate monitors

Devices occasionally record temporarily obvious faulty values ("spikes", like HR 250, cadence 300, speed 200mph etc.). Be aware, few such spikes will be "wiped out" in the "overall" average calculations, their effect on averages usually is minimal (if at all), from the mathematical point of view they do not justify an additional smoothing.

For more concrete recommendations, see here:
Elevation => Elevation FAQ
CHnuschti
 
Posts: 267
Joined: Tue Feb 13, 2007 3:28 am
Location: Switzerland


Return to FAQ

Who is online

Users browsing this forum: No registered users and 1 guest

cron