Table 1 Calculation of Bayes factors for non-significant results

From: Self-injury, suicidality and eating disorder symptoms in young adults following COVID-19 lockdowns in Denmark

Outcomes

Calculation of Bayes factor

Bayes factor

Longitudinal data

Non-significant results

Suicide attempt in women

5066^((0–1)/2) × (22/21.99644690)^(5066/2)

0.021a

Symptoms of EDs in men

2531^((0–1)/2) × (34.5/34.45555117)^(2531/2)

0.10b

Repeated cross-sectional data

Non-significant results

Self-injury

24625^((7–9)/2) × (2138.5601/2138.2811)^(24625/2)

0.00020c

Suicide ideation

24625^((7–9)/2) × (4213.1621/4212.7241)^(24625/2)

0.00015c

Suicide attempt

24625^((7–9)/2) × (167.7110/167.6976)^(24625/2)

0.00011c

Symptoms of EDs

24625^((7–9)/2) × (1859.6843/1859.5532)^(24625/2)

0.000097c

  1. Interpretations as in Jeffreys39
  2. aVery strong evidence for H0 compared to Ha.
  3. bSubstantial evidence for H0 compared to Ha.
  4. cDecisive evidence for H0 compared to Ha.
  5. Formula for Bayes factor: n^((k0–ka)2) × (RSS0/RSSa)^(n/2), where H0 = model (unweighted) without lockdown variable Ha = model (unweighted) with lockdown variable, n = size of study population, k = number of parameters in model, RSS = residual sum of squares36,37,38. Bayes factors were based on the unweighted models because the weights did not change the results notably.
Back to article page