csu-hmc / perturbed-data-paper

A paper on an elaborate gait data set.
https://peerj.com/articles/918/
11 stars 6 forks source link

Note the issues associated with the TimeStamp column in the data. #170

Closed moorepants closed 9 years ago

moorepants commented 9 years ago

I just discovered this: https://github.com/csu-hmc/GaitAnalysisToolKit/issues/133 and it would be ideal if we could mention this in this paper.

moorepants commented 9 years ago

@tvdbogert @spinningplates Do you think I need to edit the paper to reflect this issue?

tvdbogert commented 9 years ago

Does it affect the results that were presented in the paper? If not, you can just update the software so the problem does not happen again when people use the code. I suspect that the graphs in the paper won't be affected by this problem.

moorepants commented 9 years ago

Yeh, the graphs would not change in any human perceptible way because I used a trial that had no marker drop out or time stack up issues. I'm just worried that the explanation of the TimeStamp column in the data is less than informative:

  \item[TimeStamp] The monotonically increasing computer clock time when D-Flow
    receives a frame from Cortex. These are recorded approximately at
    100~\si{\hertz} sampling rate and given in seconds.

I could add a sentence or so there when the proof comes through to mention the stack up issue. Maybe that is sufficient.

tvdbogert commented 9 years ago

The explanation of the TimeStamp column is technically still correct: "when D-Flow receives a frame". I would add something like: because of buffering, we prefer derive sample times from the FrameNumber rather than TimeStamp".

The reader can then determine for themselves how much they want to trust the files that came from the Record module (which have no framenumber, right?). We should not undermine the trust in the data at the proof stage, because the reviewers (if they ever read the paper again) would feel deceived because they never had that information when they evaluated the work.

A determined and creative user can get around these issues. But it is scary that we did not even notice it for a long time, so a gentle warning in the TimeStamp explanation is needed.

moorepants commented 9 years ago

I'll mention the error to the editors, suggest the additional warning, and see what they say. They may need to send it back to the academic editor to get reviewer approval.