A multivariable approach for risk markers from pooled molecular data with only partial overlap

BMC Medical Genetics

Table 1 Top 10 SNPs according to IFs for the full data analysis resembling the “truth” (IF_full) in decreasing order

SNP		blackIF _full	blackIF _red	blackIF _part	p-value
rs7039441	✓	0.68	0.65	0.55	0.05
rs1323398	✓	0.55	0.63	0.49	0.02
rs3793482	✘	0.44	0.39	0.36	0.02
rs1048251	✓	0.38	0.27	0.40	0.09
rs10965030	✘	0.28	0.18	0.18	0.07
rs10491695	✘	0.25	0.50	0.50	0.26
rs3750417	✘	0.22	0.10	0.09	0.06
rs7846927	✓	0.21	0.12	0.19	0.05
rs6477107	✓	0.19	0.21	0.24	0.02
rs12684584	✘	0.19	0.21	0.18	0.34

✓ SNP present in both studies
✘ SNP present in the large study (study A) but not in the white✘ small study (study B)
We additionally report the respective IFs from the reduced (IF_red) and partial analysis (IF_part). Numbers are marked in bold if the IF of the (reduced or partial) analysis is smaller than that of the full analysis (IF<IF_full) and in italics if it is greater (IF>IF_full). In 13 cases, we have IF<IF_full, in 7 we have IF>IF_full. We further report p-values from univariate logistic regression for each SNP

ISSN: 1471-2350