Table 2 Random effects meta-analysis for associations of childhood predictors with belief in life after death in adulthood, and sensitivity to unmeasured confounding.

From: Childhood predictors of belief in life after death across 22 countries

Variable

RR (95% CI)

Estimated proportion of effects by threshold

I2

Global p-value

E-valuesb

< 0.90

> 1.10

Effect estimatec

95% CId

Relationship with mother when growing up

< 0.001

  

 Ref: Very bad/somewhat bad

–

–

–

–

 

–

–

 Very good/somewhat good

1.03 (1.00,1.07)

0.00

0.05

28.1

 

1.21

1.00

Relationship with father when growing up

< 0.001

  

 Ref: Very bad/somewhat bad

–

–

–

–

 

–

–

 Very good/somewhat good

1.02 (0.99,1.05)

0.00

0.05

48.0

 

1.18

1.00

Marital status of parents around age 12

< 0.001

  

 Ref: Parents married

–

–

–

–

 

–

–

 No, divorced

0.99 (0.96,1.02)

0.05

0.00

41.3

 

1.13

1.00

 Single, never married

0.95 (0.89,1.01)

0.36

0.05

77.4

 

1.29

1.00

 No, one or both had died

1.01 (0.96,1.06)

0.05

0.14

64.3

 

1.11

1.00

Subjective financial status of family around age 12

< 0.001

  

 Ref: Got by

–

–

–

–

 

–

–

 Lived comfortably

1.00 (0.97,1.03)

0.00

0.00

75.7

 

1.04

1.00

 Found it difficult

1.00 (0.98,1.02)

0.00

0.00

32.5

 

1.04

1.00

 Found it very difficult

0.99 (0.96,1.01)

0.00

0.00

< 0.1ǂ

 

1.13

1.00

Experienced abuse when growing upa

< 0.001

  

 Ref: No

–

–

–

–

 

–

–

 Yes

1.07 (1.02,1.11)

0.00

0.33

85.9

 

1.33

1.16

Felt like an outsider in the family when growing up

< 0.001

  

 Ref: No

–

–

–

–

 

–

–

 Yes

1.03 (1.00,1.06)

0.00

0.14

62.1

 

1.21

1.02

Self-rated health when growing up

< 0.001

  

 Ref: Good

–

–

–

–

 

–

–

 Excellent

1.02 (0.98,1.06)

0.05

0.09

83.0

 

1.17

1.00

 Very good

1.01 (0.98,1.04)

0.05

0.09

73.9

 

1.09

1.00

 Fair

1.01 (0.98,1.04)

0.05

0.05

54.4

 

1.10

1.00

 Poor

1.02 (0.95,1.08)

0.00

0.27

57.5

 

1.15

1.00

Immigration status

< 0.001

  

 Ref: Born in this country

–

–

–

–

 

–

–

 Born in another country

1.00 (0.92,1.08)

0.27

0.23

79.7

 

1.07

1.00

Frequency of religious service attendance around age 12

< 0.001

  

 Ref: Never

–

–

–

–

 

–

–

 At least 1/week

1.39 (1.25,1.55)

0.00

0.77

96.1

 

2.13

1.80

 1–3/month

1.26 (1.15,1.38)

0.00

0.59

93.0

 

1.83

1.56

 < 1/month

1.11 (1.05,1.18)

0.05

0.45

78.9

 

1.47

1.28

Year of birth

< 0.001

  

 Ref: 1998–2005; age 18–24

–

–

–

–

 

–

–

 1993–1998; age 25–29

1.03 (1.00,1.06)

0.00

0.09

52.0

 

1.20

1.00

 1983–1993; age 30–39

1.04 (0.99,1.09)

0.05

0.32

87.8

 

1.24

1.00

 1973–1983; age 40–49

1.05 (1.00,1.10)

0.09

0.27

88.5

 

1.27

1.00

 1963–1973; age 50–59

1.05 (0.99,1.12)

0.09

0.32

89.1

 

1.29

1.00

 1953–1963; age 60–69

1.00 (0.93,1.09)

0.23

0.32

90.5

 

1.07

1.00

 1943–1953; age 70–79

0.93 (0.84,1.04)

0.41

0.27

88.6

 

1.35

1.00

 1943 or earlier; age 80+

1.03 (0.89,1.19)

0.23

0.41

90.8

 

1.19

1.00

Gender

< 0.001

  

 Ref: Male

–

–

–

–

 

–

–

 Female

1.09 (1.03,1.15)

0.05

0.41

97.3

 

1.40

1.19

 Other

0.36 (0.08,1.71)

0.33

0.39

99.9

 

4.99

1.00

  1. aMeta-analysis for experienced abuse when growing up only includes results for 21 countries because this question was not used in Israel. All p-values passed the Bonferroni-corrected threshold of p < .0045 (p = .05/11). CI = confidence interval, RR = risk ratio. Estimated proportion of effects by threshold: The estimated proportion of effects across the implied distribution of effects sizes within the countries that are either less than a risk ratio of 0.9 or above a risk ratio of 1.1. I2: An estimate of the percent of variability in effect sizes that cannot be attributed to sampling variability. Global p-value: A test of the null hypothesis that there is no association between the predictor and afterlife belief in adulthood in all the countries. Additional heterogeneity metrics of the effects are provided in Figures S1-S27, along with the associated Q-statistics for the heterogeneity in the effects on the log(RR) scale.
  2. bThe formula for calculating E-values can be found in VanderWeele and Ding (2017).
  3. cThe E-value for the effect estimate is the minimum strength of association (on the risk ratio scale) that an unmeasured confounder would need to have with both the predictor and the outcome to entirely explain away the observed association between them, conditional on the measured covariates.
  4. dThe E-value for the limit of the 95% confidence interval closest to the null denote the minimum strength of association (on the risk ratio scale) that an unmeasured confounder would need to have with both the predictor and the outcome to shift the confidence interval to include the null value, conditional on the measured covariates.
Back to article page