Mental Health

e Switch: employment status moderation

f PS5: age moderation Fig. 3 | Machine learning results illustrating estimated CLATEs $( N = 8 , 1 9 2 )$ .

a Switch: gender moderation

c Switch: household structure moderation

d PS5: household structure moderation

f PS5: employment status moderation Fig. 4 | Machine learning results of effect modification contrasting ownership of Switch and PS5 $( \pmb { N } = \pmb { 8 } , \pmb { 1 9 2 } )$ . The IV causal forest was used to estimate CLATEs of game console ownership on K6. Each plot is a heat map of a trivariate distribution where the colour gradient visualizes the average value of z within y-axis and x-axis bins of rectangular fields. The estimated CLATEs of video game console ownership on K6 are illustrated, with age on the y axis and each background characteristic on the x axis. A lighter shade indicates that video game ownership is more advantageous for individuals in that particular bin. The bar width of the heat maps represents the sample size for each age group. The estimates are standardized by the s.d. a,b, Gender moderation for Switch (a) and PS5 (b). c,d, Household structure moderation for Switch (c) and PS5 (d). e,f, Employment status moderation for Switch (e) and PS5 (f).