There is something to it, unfortunately I only have data on runs.
This week I didn’t feel so well, not sure what it was. This is a 90min run on Thursday. Following a rest day (I always feel sluggish after rest days). Heart rate and pace would indicate around AeT. However, this shows clearly a more demanding effort:
Furthermore, I’ve started tracking morning HRV again (wanted to give it a try again). The days around the more demanding run would show a fairly low morning rMSSD reading, while the “normal feeling” run had expected rMSSD values.
Definitely something to play with for the next few weeks.
Are just running the python from collab? My Garmin 530 has been logging HRV and I’d like to run some data thru it, but not on Google servers. So I’m thinking of pulling out the code and running it on my computer.
Thought I’d post my n=1 recordings.
Perhaps the best one I have is from a cardiac stress test I did on the trainer.
Visually it tells me one thing, numerically another; so…I can only guess at the efficacy.
Inflection point @ 115bpm with DFA/alpha 1 @ 1.18…
What I did in my initial testing was split the RR data into 1 min intervals and plot the DFA alpha-1 (under the nonlinear tab on kubios) against time/hr/power. Maybe try that instead as you’re currently just looking at the raw RR intervals
Hi Marco,
Very thanks for this tool first!
Although I still can’t read data from that, but I think that’s because I use wahoo .fit file it’s include much data(power meter, Garmin 945 HR, Garmin 945 GPS).
Dose this tool only for HRV data?
If that’s true, How can I get HRV form Garmin 945?
if possible please help me, I really interesting in this issue.
Similarly to you, I had a bad day yesterday. Slept 9.5 hours but running too fast and doing a SS session on Saturday showed. Before this cycling session I did an hour long run which was pure suffering (legs didn’t work). It felt better on the trainer but the graph shows it clearly wasn’t AeT (Phoenix -2, IF 0.75):
thanks! Indeed unfortunately there are some requirements in terms of the format of the data and not all devices collect beat to beat data (averaged heart rate cannot be used for this specific analysis, even though it can provide you with useful insights on cardiac decoupling, another method that can help identifying the aerobic threshold).
certainly agree with you. I believe these methods tend to highlight other aspects linked to how cardiac responses change wrt to exercise intensity and also breathing, which might be indicative of different thresholds despite the fact that there isn’t any parasympathetic activity to measure. In my experience, looking at measures of parasympathetic activity such as rMSSD indeed provide values close to zero no matter the intensity, while other features such as alpha 1 used here, have different properties and could be indicative of the mechanisms above. Hope this makes sense to you!
Hi, I am one of the authors of the Frontiers DFA a1 “perspective review”. For those interested I have quite a few blog posts about the subject, answers to many questions and some guides on how to implement the various methods (constant power vs ramps). As far as the issue of running vs cycling, we have not seen any disparity in DFA a1 behavior, they should be equivilant. muscleoxygentraining.com/2020/05/dfa-a1-vs-intensity-metrics-via-ramp-vs.html
My conclusion after playing around with it a little bit further: interesting for sure but it is still too variable and complicated. There are many interesting studies cited in the Frontiers review but when it comes to the regular athlete it’s to variable and complicated (yet). I find the marker “upper zone 2 is where breathing speeds up and talking becomes more difficult” more reliable and easier to determine. Far easier.
My main issue with this alpha1 analysis is the pre-processing of the data. Identification of artefacts and filtering of the data. You need a lot of experience with this. And depending on what you apply you get very different results.
So even for very steady “real world” efforts you get quite some variation:
quite frankly, “breathing speeds up, talking gets slightly more difficult” does a better job for me. However, I would not discard the entire approach. We’re simply not there yet.
Right. You can bend your data any way you want and then you need backtesting to see if your massage helped to reduce signal to noise ratio or you actually amplified some artifacts. Testing here means training according the estimated intensity and looking at the performance. Which you can of course do even without HRV.
I agree with you but I’ll play with it some more to see if it actually helps with identifying some overreaching and such (I can tell how I feel but a second metric to confirm would be nice). I don’t have many observations yet, my HRV was turned off until Sunday unfortunately