Closed trvrb closed 11 years ago
I'm not sure what, exactly, to do here. This figure would normally be a supplementary figure, but eLife doesn't do supplementary figures. This definitely seems to fit with "validation" which is this section.
Reviewer 1 is rather supportive, so you could argue to keep it in, indeed because there is no opportunity to put it in an SI for example. But a simple alternative would be to support the statement that log$_2$ HI titer is independent of antigenic distance, and also that absolute error is independent of time, with a correlation coefficient in the text and remove the figure. I asked some stats advice from Marc on this and he says that if we need to also rule out non-linear relationships, one could bin the variables into discrete outcomes and use a chi-squared test (if binned "properly" non-linear effects come out as well). He also pointed at this link for a list of tests: http://stubber.math-inf.uni-greifswald.de/~vollmer/ERCIM11.pdf.
I like this idea. I might start with just Spearman rank correlations however.
We agree and have revised the manuscript to remove this figure. Instead of the figure we report Pearson’s correlations between antigenic distance and error ( r = 0.098 ) and between year and error ( r = −0.007 ).