Data analysis with Shapley values for automatic subject selection in Alzheimer’s disease data sets using interpretable machine learning

Table 2 Demographic data, cognitive test scores, and the number of ApoE ε4 alleles of the selected AIBL subjects separated by diagnosis group

Variable	sMCI	pMCI	Σ	p value
n	16	12	28
Age (in years)	77.8 ±6.9	75.3 ±5.8	76.8 ±6.5	0.2856
Gender (proportion of females)	37.5%	33.3%	35.7%	1.0000
Gender (proportion of males)	62.5%	66.7%	64.3%
MMSE	28.0 ±1.7	26.3 ±1.6	27.3 ±1.8	0.0133
CDR	0.4 ±0.2	0.5 ±0.0	0.5 ±0.1	0.2317
ApoE ε4 (proportion of subjects with 0 alleles)	56.3%	16.7%	39.3%	0.0550
ApoE ε4 (proportion of subjects with 1 allele)	37.5%	50.0%	42.9%
ApoE ε4 (proportion of subjects with 2 alleles)	6.3%	33.3%	17.9%
Time to final diagnosis in months	41.3 ±22.4	28.4 ±14.2	35.7 ±20.1	<0.2457

For continuous features, mean and standard deviation are given. p values are calculated using Mann-Whitney U test [49, 50] for continuous features and using χ²-test for ordinal and nominal features

ISSN: 1758-9193