Data handling query

[Tom M]

Have got some data that is best expressed as the percentage change from baseline. The shot below shows the raw data (in this case heart rate) at baseline, then at subsequent timepoints after a period of exercise. In order to express this data as percentage change from baseline I have two options it seems.

Method one (green shading) is what I would be more familiar with whereby each individual is normalised to their own baseline. For example, for subject one at time point 60, contents of cell G2 (73.64) is 15.99% greater than their baseline in B2 (63.49).

Method two (blue shading) is where each individual is normalised to the mean group baseline. In this case, again for subject one at time point 60, using this method, contents of cell G2 (73.64) are 6.75% greater than the group mean baseline in cell B18 (68.98).

Still awake?

Looking at it on an individual basis, it seems a bit oddd, but am more interested in the group data as a whole, which is pretty similar between the two methods. Method two was suggested by someone to be more appropriate, but never really got a satisfactory answer why. Method one does yield smaller errors (good), but doesn't allow you to express variability at baseline in the way method two does, which I quite like.

Anyway, any thoughts from the collective branes?

Thanks.

[mrcharly-YHT]

For method 2 do you apply the same normalisation factor to each individual's raised heartrate?

[tonycollinet]

Method 1 will show you how individuals normalised CHANGE in heartrate will vary within the population.

Method 2 to me is mixed up. First you are normalising to a common "average" baseline, and then looking at individual change in heartrate relative to a common baseline. The figures in this case mix two sets of data - first is the absolute variation in resting heartrate compared to the population average, and then the variation in how heartrate changes amongst individuals within the population. I don't see how this can be useful.

[David Martin]

Method 2 is wrong. It just is.

If you want to see how individuals compare to the group  in terms of their response, first normalise the data to get the proportional change for the individual, then calculate how those proportional changes vary for the individual.

You might want to try normalising to the HR range for each individual as a proportion of HRmax-baseline instead of just proportion of baseline. An individual with a RHR of 50 and max of 190 will have much more capacity than an individual with RHR of 80.

You could also try absolute numerical change. But normalise inside each variable (individual) before doing the group comparison.