Statistical analyses
The correlation between tarsus length and telomere length
We first tested the phenotypic correlation between TL and tarsus length (as a proxy for body size) within 2462 house sparrow nestlings from Hestmannøy and Træna. TL (response variable) was log10-transformed and linear mixed-effects models (LMMs) were fitted with a Gaussian error distribution (R package ‘lme4 ’, Bates, Mächler, Bolker, & Walker, 2015). Sex differences in TL are known for house sparrows (Pepke et al., 2021,submitted ). Thus, models included sex, (continuous) fledgling age at sampling, hatch day (ordinal date mean centered across years), and island identity as fixed effects. We fitted random intercepts for brood identity and year to account for the non-independence of nestlings from the same brood and year. Because our study populations are known to be affected by inbreeding depression (Niskanen et al., 2020), we included the inbreeding coefficient (F ) as a fixed effect (Reid & Keller, 2010). We then compared models with and without (age-standardized) tarsus length using Akaike’s information criterion corrected for small sample sizes (AICc , Akaike, 1973; Hurvich & Tsai, 1989), and Akaike weights (w ) and evidence ratios (ER ) to determine the relative fit of models given the data (Burnham & Anderson, 2002). Models were validated visually by diagnostic plots and model parameters are from models refitted with restricted maximum likelihood (REML). Estimates and 95% confidence intervals (CI) are reported.