Table 2 Explaining social isolation.

From: Using machine learning to understand social isolation and loneliness in schizophrenia, bipolar disorder, and the community

Predictor Variable

β Value

% of model runs retained

Main Effects

 Loneliness

0.11

100

 Social Anhedonia

0.21

100

 Social Avoidance

 

41.03

 Depression

 

62.82

 Nonsocial Cognition

−0.06

80.77

 Social Cognition

 

55.12

Group Interactions

 Loneliness x Bipolar

0.04

71.72

 Loneliness x Community

0.18

97.43

 Social Anhedonia x Bipolar

 

55.13

 Social Anhedonia x Community

 

39.74

 Social Avoidance x Bipolar

0.10

78.21

 Social Avoidance x Community

 

64.10

 Depression x Bipolar

 

26.92

 Depression x Community

 

44.87

 Nonsocial Cognition x Bipolar

0.05

70.51

 Nonsocial Cognition x Community

 

55.13

 Social Cognition x Bipolar

 

60.25

 Social Cognition x Community

 

46.15

  1. LASSO regression model of schizophrenia, bipolar disorder, and community samples (R2 = 0.33). Schizophrenia was used as the reference group in all interactions. An empty β value indicates no or practically no independent contribution of a predictor variable to social isolation above and beyond other variables. Presented on the right is the percentage of runs in which a predictor variable was retained in the model (i.e., its β value not shrunk to 0). The higher the percentage, the more robust the variable’s contribution.
  2. LASSO Least Absolute Shrinkage and Selection Operator.
  3. Variables in bold were retained in the final model.
Back to article page