When running time-lagged segment analyses in R the outcome variable is significantly predicted by the predictor variable at t-1 but not by the predictor variable at t, which seems not logical to me. Actually, the predictor variable at t-1 is only added to the model to be controlled for.