Using Bonferroni, BIC and AIC to assess evidence for alternative biological pathways: covariate selection for the multilevel Embryo-Uterus model

Table 6 AIC (difference from null model) for the motivating dataset

Variable	AIC
Variable	null	E	U	E + U
Number of embryos transferred	0	−0.6	1.1	−0.1
Age group	0	−178.2	−153.8	−175.1
Number of embryos created	0	−13.7	−19.3	−9.3
IVF Attempt number	0	−4.3	−8.1	−2.6
ICSI¹	0	0.5	1.6	2.4
Pregnancy History²	0	−2.6	−8.5	−3.1
Duration infertile	0	−1.7	−10.4	−3.0
Tubal diagnosis	0	−9.9	−14.2	−12.8
PCO³ diagnosis	0	2.0	2.0	3.9
Endometriosis	0	0.1	−1.2	0.8
Idiopathic diagnosis	0	1.8	1.6	2.4
Male diagnosis	0	0.6	1.3	2.6
Donor sperm	0	2.0	0.8	1.1
Transfer day⁴	0	−3.5	−1.9	−1.6
Year	0	−8.0	−11.2	−13.0
Treatment Centre	0	−11.2	−9.2	−24.7

¹Intercytoplasmic sperm injection, a variation on standard IVF.
²A composite variable indicating the number and type of previous pregnancies.
³Polycystic Ovary Syndrome.
⁴Embryos were cultured for 2 or 3 days before transfer to the potential mother.
AIC (difference from null model) for models for the motivating dataset in which each included variable is assigned to either none (null), the embryo (E), uterus (U) or both (E + U) sub-models. The highlighted entries indicate the assignment in the final selected model, usually (but not in all cases – see text) that with the lowest AIC.

ISSN: 1471-2288