Table 5:

Comparison of the LASSO estimates selected from the NSW-NSW dataset and the NSW-PSID dataset by minimizing the validation errors. For each validation method, we compute ${\Vert {\widehat{\tau }}_{NSW}-{\widehat{\tau }}_{PSID}\Vert }_{2,t}:={\left(1/n_t{\sum }_{W_i=1}{\left({\widehat{\tau }}_{NSW}-{\widehat{\tau }}_{PSID}\right)}^2\right)}^{1/2}$ – the treatment group L₂ norm of the difference of the selected LASSO estimates ${\widehat{\tau }}_{NSW}$ from the NSW-NSW dataset and ${\widehat{\tau }}_{PSID}$ from the NSW-PSID dataset. The results regarding the L₂ norm based on the NSW-NSW control group and the NSW-PSID control group are similar and are presented in Table 6 in the appendix.

method	prd	cvr	full	S-M	prgn	combo
${||{\widehat{\tau }}_{\mathrm{NSW}}-{\widehat{\tau }}_{\mathrm{PSID}}||}_{2,\text{t}}$ × 102	0.74	1.36	0.00	1.36	4.36	0.00