The paper that convinced me GRFs *could* be used to predict tibia bone load (TBL):
“Ground reaction force metrics are not strongly correlated with tibial bone load when running across speeds and slopes...” by @EmilyMatijevich @KarlZelik
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0210000
Why? Well... 1/4
“Ground reaction force metrics are not strongly correlated with tibial bone load when running across speeds and slopes...” by @EmilyMatijevich @KarlZelik
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0210000
Why? Well... 1/4
There was no strong (r>=0.8) correlation between peak vGRF & peak TBL when all the slopes/speeds were combined & correlations were averaged across subjects (r=0.72 +/- 0.42).
Paper data is open source, so what do the correlations for each slope look like? Pretty strong:
Paper data is open source, so what do the correlations for each slope look like? Pretty strong:
So let’s try to predict peak TBL from peak vGRF, speed, & slope! I fit a *basic* linear regression to 9 subjects in the paper (peak TBL = 0.23 + 2.96*peakGRF + 0.12*speed + 0.10*slope) & tested on the 10th. Model predicts peak TBL across speeds/slopes with an RMSE = 0.37 BW
:

All this to say "Don’t give up on GRFs"!
GRF metrics might not strongly correlate w/ tibia bone load across speeds/slopes, but a simple regression can predict peak tibia bone load from peak GRF *if* you just include speed/slope.
Data/code for plots: https://gist.github.com/alcantarar/45dbf96fc8a7d3f4cbf0b003eb38ec46
GRF metrics might not strongly correlate w/ tibia bone load across speeds/slopes, but a simple regression can predict peak tibia bone load from peak GRF *if* you just include speed/slope.
Data/code for plots: https://gist.github.com/alcantarar/45dbf96fc8a7d3f4cbf0b003eb38ec46