Continuous NHANES Web Tutorial: Clean & Recode Data: Task 2c

Task 2c: How to Identify and Recode Skip Patterns Using Stata

The second task is to check the data for skip patterns. To do this, you will use the:

sample person questionnaire, and
Stata commands to get cross tabulations tables, and then
recodeas necessary.

Step 1: Check codebook for skip patterns

Check the codebook to determine if a skip pattern affects the variables in your analysis. See the Locate Variables module Task 1 for more information on how to locate background information on variables in the documentation.

Skip Pattern in Blood Pressure Questionnaire Codebook

screenshot of Skip Pattern in Blood Pressure Questionnaire Codebook

Step 2: Check data for skip patterns

After you have used the codebook to discover if a skip pattern affects variables in your analysis, you will use cross tabulations obtained by the Stata tabulate command to determine the presence of skip patterns.

Program to Check Data for Skip Patterns

Use the tabulate command to determine the frequency of each value of the variables listed for participants who were interviewed and examined in the MEC and who were age 20 years and older. Use the missing option to display the missing values. Two variables listed on the tabulate command line will create a crosstab with BPQ020 as the row variable and BPQ030 and BPQ050a as the column variables.

tabulate bpq020 if (ridageyr >=20 & ridageyr <.) & ridstatr==2, missing
tabulate bpq030 if (ridageyr >=20 & ridageyr <.) & ridstatr==2, missing
tabulate bpq050a if (ridageyr >=20 & ridageyr <.) & ridstatr==2, missing
tabulate bpq020 bpq030 if (ridageyr >=20 & ridageyr <.) & ridstatr==2, missing
tabulate bpq020 bpq050a if (ridageyr >=20 & ridageyr <.) & ridstatr==2, missing

Highlighted items from the tabulate output for skip patterns:

Notice the high number of missing values (n=6,646) for BPQ.030 compared to the much lower number of missing values in the prior question BPQ.020 (n = 106).
Further down, the output includes a cross tabulation of BPQ.020 responses by BPQ.030 responses. Note the large number of missing values in BPQ.030 (n=6,540) for those who responded with a "No," coded as "2," for BPQ.020. These respondents were not asked because of a skip pattern and therefore, these responses will need to be recoded before the data are further analyzed.