The main dropout-contagion findings remained essentially unchanged across five robustness checks including unmatched samples, IPTW, and conditional regression¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r2 · q2
Currency · newest study 2021 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Rerunning the regression models on the full sample including unmatched controls and on exposures after the risk index yielded stronger effect sizes, consistent with matching yielding conservative estimates. → Dupéré 2021 q2 i? Rerunning the models with risk-index items as separate variables, with IPTW, and with conditional logistic regression left the main findings essentially unaffected. → Dupéré 2021 (2)

Evidence¶

Dupéré 2021¶

Dupéré, V., Dion, E., Cantin, S., Archambault, I., & Lacourse, E. (2021). Social contagion and high school dropout: The role of friends, romantic partners, and siblings. Journal of Educational Psychology, 113, 572-584. https://doi.org/10.1037/edu0000484

q2 · i? · associational · r2

Robustness tests reran the regressions on the full sample (N = 545) including unmatched controls, and on exposures occurring after administration of the risk index to probe reverse causality. The article reports "the effect sizes and significance levels tended to be stronger, a result consistent with scholarship showing that matching tends to yield conservative estimates".

"In both tests, the effect sizes and significance levels tended to be stronger, a result consistent with scholarship showing that matching tends to yield conservative estimates"

Dupéré 2021 (2)¶

Dupéré, V., Dion, E., Cantin, S., Archambault, I., & Lacourse, E. (2021). Social contagion and high school dropout: The role of friends, romantic partners, and siblings. Journal of Educational Psychology, 113, 572-584. https://doi.org/10.1037/edu0000484

q2 · i? · associational · r2

Alternative confounder-handling specifications — risk-index items entered separately and inverse probability of treatment weighting, after which exposed and non-exposed participants were successfully rebalanced — left the main findings stable; conditional logistic regression also left "the main findings were essentially unaffected".

"With both tests, the main findings were essentially unaffected."

Discussion¶

Related Claims¶

• Dropout among friends, romantic partners, and siblings each shows an independent association with an adolescent's own dropout — related
• Adolescents are at greater risk of dropping out when at least one same-age intimate (friend, romantic partner, or sibling) has recently dropped out of school — related
• The association between network dropout and adolescents' own dropout does not significantly differ by gender or by regular versus special education sector — related
• Risk of dropout is especially high when an adolescent's network includes more than one type of same-age intimate who recently dropped out — related
• Only recent (within the past year) occurrences of dropout in an adolescent's network are associated with the adolescent's own dropout; older occurrences are not — related