@kylehamilton
Looking at the cache_dataset script, it seems like no correlation can be imputed if there is no correlation already coded or calculated from t, means and SD. Because the imputation uses Hmisc::impute(fun = "random") we're just relying on some correlation being present. We discovered that if we run the function on a correlation column that only contains NA, the function gives an error. However, the cross-situational word learning dataset still calculates d_var_cal, for within_two studies that do not have correlation values.
@kylehamilton Looking at the cache_dataset script, it seems like no correlation can be imputed if there is no correlation already coded or calculated from t, means and SD. Because the imputation uses Hmisc::impute(fun = "random") we're just relying on some correlation being present. We discovered that if we run the function on a correlation column that only contains NA, the function gives an error. However, the cross-situational word learning dataset still calculates d_var_cal, for within_two studies that do not have correlation values.