Algorithm 1 Elastic Net LOSO pipeline.

Require: Dataset $D={\left\{({\mathbf{x}}_i,y_i,s_i)\right\}}_{i=1}^n$, participant set $\mathcal{S}$, candidate features $\mathcal{F}$   1:for all $s\in \mathcal{S}$ do   2:      $D_{\mathrm{train}}\leftarrow \{({\mathbf{x}}_i,y_i,s_i)\in D:s_i\ne s\}$   3:      $D_{\mathrm{test}}\leftarrow \{({\mathbf{x}}_i,y_i,s_i)\in D:s_i=s\}$   4:      ${\mathsf{\mu }}_s\leftarrow$ column means of $D_{\mathrm{train}}\left[\mathcal{F}\right]$   5:      $(X_{\mathrm{train}},X_{\mathrm{test}})\leftarrow (D_{\mathrm{train}}\left[\mathcal{F}\right],D_{\mathrm{test}}\left[\mathcal{F}\right])$ with missing entries replaced by ${\mathsf{\mu }}_s$   6:      ${\mathsf{\sigma }}_s\leftarrow$ column standard deviations of $X_{\mathrm{train}}$   7:      $X_{\mathrm{train},f}\leftarrow (X_{\mathrm{train},f}-{\mu }_{sf})/{\sigma }_{sf}$ for every $f\in \mathcal{F}$   8:      $X_{\mathrm{test},f}\leftarrow (X_{\mathrm{test},f}-{\mu }_{sf})/{\sigma }_{sf}$ for every $f\in \mathcal{F}$   9:      $y_{\mathrm{train}}\leftarrow$ outcome vector from $D_{\mathrm{train}}$ 10:      $M_s\leftarrow \mathtt{ElasticNetCV}(X_{\mathrm{train}},y_{\mathrm{train}},\mathrm{cv}=5)$ 11:      ${\widehat{y}}_s\leftarrow M_s\left(X_{\mathrm{test}}\right)$ 12:      ${\widehat{\mathsf{\beta }}}_s\leftarrow$ coefficient vector of $M_s$ 13:      $z_{sf}\leftarrow 1$ if ${\widehat{\beta }}_{sf}\ne 0$, and $z_{sf}\leftarrow 0$ otherwise, for every $f\in \mathcal{F}$ 14:end for 15:return held-out predictions, coefficients, and non-zero-coefficient indicators