Data cleaning

To have meaningful test results, we need to clean the datasetnhanes.nh1516 first. In the dataset, we saw the records with "refused to answer" or "do not know" or else.

Usually we need to recode those records as missing because in the dataset they are treated as a category, but they are no real meaning.

To do that, runprocfreq to check which variables have those codes and need to recode .

For example ,

/***DUQ200: Ever used marijuana or hashish***/

procfreqdata = nhanes.nh1516;table DUQ200;run;

You can see three people answered 'refused' (7) and three people answered 'Do not know ' (9 ) once you check the dictionary from the CDC website. We will recode those people as missing.

Ever used marijuana or hashish Cumulative Cumulative DUQ200 Frequency Percent Frequency Percent 1715 50.03 1715 50.03 2 1707 49.80 3422 99.82 3 0.09 3425 99.91 9 3 0.09 3428 100.00 Frequency Missing = 6543