6

Background I am comparing data on how ATPase activity from flight and leg muscle from locusts reacts to different concentrations of calcium. I plotted a scatter graph of each data set on the same graph (ATPase activity on the Y axis, calcium concentration on the X axis) and have a trend line for each. The trend line for leg muscle has a steeper gradient, suggesting it is more responsive to changes in calcium.

I carried out 9 repeats for each muscle type at each concentration of calcium.

Question How do I test if there is a statistically significant difference between the trends of each data set?

I am a novice user of R and excel, though any general pointers would be appreciated.

D Greenwood
  • 160
  • 1
  • 1
  • 6
  • 5
    Hi D Greenwood, welcome. These three posts from the site could be helpful: [First](http://stats.stackexchange.com/questions/55501/test-a-significant-difference-between-two-slope-values), [second](http://stats.stackexchange.com/questions/33013/what-test-can-i-use-to-compare-slopes-from-two-or-more-regression-models?rq=1), [third](http://stats.stackexchange.com/questions/59327/how-to-compare-two-regression-slopes-for-one-predictor-on-two-different-outcomes/59331#59331). Since you have repeated measurements, you'll probabily need to fit a linear mixed effect model (e.g. with `lme` in `R`). – COOLSerdash May 23 '13 at 18:53
  • Thanks CoolSerdash for the links. However the trends are not linear and the data collected at each calcium concentration does not follow a normal distribution. I was thinking of carrying out a Mann-Whitney u test between the two data sets for each Calcium value individually but this seemed a bit long winded. – D Greenwood May 23 '13 at 19:06
  • 1
    Could you maybe post your models (and scatterplots)? You say that the trends are not linear. How did you fit the trends? Are they polynomial (e.g. quadratic)? – COOLSerdash May 23 '13 at 19:42

1 Answers1

1

you can try log transforming the data on either of both axis to get a normal distribution. Then do linear regression and compare the slopes to see if they are significantly different (calculate Z and then look up p-value in statistics table).

Alternatively you can calculate area under the curve (AUC) using the trapezoidal rule, for each rep, this should give a normal distribution and then do a standard Student's t-test. You can do this for different sections of your curve.

igm
  • 11
  • 2