The starting sample comprised 13,978, and of these 7,870 (56.3 %) provided at least one measure of depressive symptoms. (1,688 respondents provided only one SMFQ measure, 1,788 = 2 SMFQ measures, 1,837 = 3 SMFQ measures, 2,557 = complete set of 4 measures). Table 1 shows strong evidence of a relationship between sociodemographic variables and missing data on depressive symptoms. The main focus for the modelling of depressive symptoms will be those with at least half of the required SMFQ data (2,942 male/3,240 female). Among these, the numbers with follow-up information on alcohol at age 18.5 was further reduced (965 male/1,660 female).Table 1Demographics against SMFQ data availability n Data availability χ 2, p No SMFQ data (n = 6,112) (%)1 measure (n = 1,682) (%)2 measures (n = 1,788) (%)3 measures (n = 1,837) (%)4 measures (n = 2,557) (%)Sex Male7,22056.847.956.847.441.3 χ 2 = 217.6, p < 0.001 Female6,75643.252.143.252.658.7Housing tenure Mortgaged/owned9,55961.273.780.083.188.2 χ 2 = 868.4, p < 0.001 Private rented1,38413.711.68.48.26.7 Subsidized rented2,08225.114.711.78.75.2Parity First born5,77041.741.643.249.050.8 χ 2 = 124.7, p < 0.001 Second born4,53934.736.437.134.634.2 Third born plus2,61823.622.019.816.415.0Maternal education A level or
