1[2 {"id":"C1","text":"Memo's conclusion: the slope of y on x is statistically significant at the 5% level.","results_keys":["memo_claim_supported","p_hac","t_hac","beta_ci_low","beta_ci_high"],"status":"unsupported","caveat":"Based on Newey-West HAC SE (Bartlett, 8 lags, normal reference); p_hac=0.118 and the 95% CI contains 0."},3 {"id":"C2","text":"The memo's point estimates and conventional SE/t/p are reproduced from the data.","results_keys":["alpha_hat","beta_hat","se_iid","t_iid","p_iid","n_analysed"],"status":"supported","caveat":"Agreement to the memo's reported rounding."},4 {"id":"C3","text":"OLS residuals are strongly positively autocorrelated, so iid SEs are inappropriate.","results_keys":["rho_resid"],"status":"supported","caveat":"rho_resid=0.61, lag-1 sample autocorrelation."},5 {"id":"C4","text":"The HAC SE is about 1.45 times the iid SE.","results_keys":["se_hac","se_iid","hac_lags"],"status":"supported","caveat":"Bandwidth from Andrews (1991) AR(1) plug-in."},6 {"id":"C5","text":"The non-significance verdict is robust to HAC bandwidths from 5 to 24 lags.","results_keys":["hac_lag_sensitivity","hac_lag_sensitivity_min_p_excl_0"],"status":"supported","caveat":"Only L=0 (no autocorrelation correction) gives p<0.05."}7]8