,
11 tweets,
4 min read
Read on Twitter
@Sean_Harrison2 Dear Sean, we have carefully considered your comments on twitter and worked to reconcile your analyses with ours. The main difference stems from the following: 1/11
@Sean_Harrison2 1. We are analyzing different markers. The marker you are analyzing (rs113010081) is in LD with delta32, but the probe for the marker we are analyzing (rs62625034) directly targets delta32. The SNP is described in the section entitled “Marker selection and validation”. 2/11
@Sean_Harrison2 We used rs113010081 as an additional validation, but all the analyses in the paper are based on rs62625034 which should be the correct marker to use. I am sorry if this was not apparent from the wording in the manuscript. 3/11
@Sean_Harrison2 2. Our filtering schemes are different. We have redone our analyses mimicking your filtering scheme while maintaining as many unrelated delta32/delta32 individuals as possible. Se results below. 4/11
@Sean_Harrison2 3. You use more covariates than us, including ‘center’. We have redone our analyses using the additional covariates you suggest. See results below 5/11
@Sean_Harrison2 4.We express mortality as a function of age rather than time since enrollment. We will argue that this is a more appropriate analysis as mortality is not a linear function of age (see also page 131-135 in book Kleinbaum and Klein). 6/11
@Sean_Harrison2 5. Inspecting your scripts, it looks like you only considered individuals who die, not individuals who survive. Ignoring overall survivorship leads to a marked reduction in statistical power. Let us know if we misunderstood. 7/11
@Sean_Harrison2 When redoing the analysis, removing up to 3rd degree kin (i.e. all pairs in ukb3367_rel_s488364.dat) (+ ppl with more than 10 kins) and controlling for sex, PCs, and center, we get a p-value of 0.015, similar to what we reported in our paper. 8/11
@Sean_Harrison2 Points 2-4 does not seem to make much difference in the calculation of the p-value in this case. The main difference is in points 1 and 5. If we redo the analysis ignoring living people and using the wrong marker we instead get a non-significant p-value like you did 9/11
@Sean_Harrison2 I will send you an email and I hope you will continue the conversation - and maybe you can correct us if our reading of your script is not correct. 10/11
@Sean_Harrison2 If you, or anybody else, have questions about the paper or the analyses, please feel free to contact us. We will be happy to help, and we apologize if some of the Methods section is difficult to parse 11/11
