I am measuring lung function using FEV1 (%) and CT score (range 0-30) in 20 patients at time 0 and at one year of follow up. How should I test which of the two measures (FEV1% or CT Score) shows greater change from time 0 to time 1 year? If I rescale CT score to a 0-100 scale and then compare the percent change in FEV1% and the rescaled CT Score, would that be appropriate?
no need to rescale CT score, just compare % change in each indicator.
eg if FEV1% goes from 70 to 75, % increase is 5/70 = 7.1% and if CT score goes from 20 to 25 (ie getting better) % increase is 5/20 = 25%.
I would additionally consider the variance of the measurements.
I extend the example given by James: If the variance of the (small) FEV1% increase is almost zero, but the variance of the (large) CT score increase is gigantic, then the FEV1% might yet be the better indicator to assess the lung function, despite its comparably small relative change.
PS:
I am not sure is a simple "percent change" is comparable, as the scales are arbitrary and the results sensitively depend on the starting value. Example: a CT score change from 0 to 1 (or any other value) would be an infinitively large relative change, from 1 to 2 it would be a 100% change, and from 29 to 30 it would be a change for only about 3%. If you rescale the scores (why not?) you will get completely different values. Similar arguments apply to FEV1%, as it makes a difference if you start with 80% or with only 60%. If there are norm values available one may calculate the change relative to that norm value. Otherwise I would use the logit-transformed values of FEV1% (in (0,1)) and of the score scaled to (0,1) and use the (absolute) changes on the logits.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS