NHANES Dietary Web Tutorial: Review Data & Create New Variables: Task 2

Task 2: How to Identify and Recode Skip Patterns in NHANES Data

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

Check the codebook for skip patterns
Check the data for skip patterns
Create new variables, as necessary.

Step 1: Check the Codebook for Skip Patterns

Check the codebook to determine whether a skip pattern affects the variables in your analysis. See Task 1 of the Locate Variables module for more information on how to use the documentation to obtain background information about your variables.

Skip Pattern in Osteoporosis Questionnaire Codebook

screenshot of skip pattern in osteoporosis questionnaire codebook

Step 2: Check the Data for Skip Patterns

After you have used the codebook to determine whether any of the variables in your analysis are part of a skip pattern, you will use the PROC FREQ procedure in SAS to obtain cross-tabulation tables using the variables that are part of the skip pattern. This will help confirm the presence of the skip pattern in the data. This example is from the "Supplement" program.

Program to Check Data for Skip Patterns

*--------------------------------------------------------------------;
* Use the PROC FREQ procedure to determine the frequency of each     ;
* value of the variables listed. Use the TABLES statement to list   ;
* the variables to be included in the output frequency table and the ;
* cross tabulation frequency table for the skip patterns. Note that ;
* an asterisk indicates that a crosstab will be constructed.         ;
*--------------------------------------------------------------------;

proc freq data=DEMOOST;
     where WTINT2YR > 0 and RIAGENDR=2 and RIDAGEYR >= 20;
      tables OSQ060*OSQ070/list missing;
      title 'Check skip pattern for osteoporosis questionnaire';
run;

Output of Program

Click here to view program output and highlights

Check skip pattern for osteoporosis questionnaire

The FREQ Procedure

Cumulative Cumulative
OSQ060 OSQ070 Frequency Percent Frequency Percent

-------------------------------------------------------------------------------

.              .                1        0.04             1         0.04
Yes            Yes             232        8.84           233         8.88
Yes            No               85        3.24           318        12.12
Yes            9                 3        0.11           321        12.24
No             .              2292       87.38          2613        99.62
9              .                10        0.38          2623       100.00

Highlighted items from the PROC FREQ output for skip patterns:

• The output includes a cross tabulation of OSQ060 responses by OSQ070 responses. Note the large number of missing values in OSQ070 (n=2,292) for those who responded with a "No" for OSQ060. These participants were not asked OSQ070 because of a skip pattern. Therefore, these responses will need to be recoded, as demonstrated in Step 3, before the data are further analyzed.

Click here to close program output and highlights

Step 3: Create New Variables, As Necessary

Using the SAS IF, THEN, and ELSE statements in a data step, you can create new variables derived from the values of the variables in the skip pattern sequence. This example is from the "Supplement" program.

For more information, see Task 4: Create New Variables.

Program to Create a New Variable

*--------------------------------------------------------------;
* Create a new variable called TREATOSTEO based on responses ;
* to the variables OSQ060 and OSQ070. ;
*--------------------------------------------------------------;

data DEMOOST;
    set DEMOOST;
    if OSQ070=1 then treatOSTEO=1;
    else if OSQ070=2 or OSQ060=2 then treatOSTEO=2;

run;

*---------------------------------------------------------------;
* Use the PROC FREQ and TABLE statements to check the derived ;
* variable (TREATOSTEO) against the original variables (OSQ060 ;
* and OSQ070). ;
*---------------------------------------------------------------;

Proc freq data=DEMOOST;
where WTINT2YR > 0 and RIAGENDR=2 and RIDAGEYR >= 20;
table TREATOSTEO*OSQ060*OSQ070/list missing;
format TREATOSTEO OSQ060 OSQ070 YESNO.;
title 'Check derived variable TREATOSTEO';
run;

Output of Program

Click here to view program output and highlights

Check derived variable TREATOSTEO

The FREQ Procedure

                                                                    Cumulative    Cumulative
treatOSTEO    OSQ060         OSQ070    Frequency      Percent        Frequency      Percent
----------------------------------------------------------------------------------------------
.              .              .                11        0.42            11            0.42
.              Yes            .                 3        0.11            14            0.53
Yes            Yes            Yes             232        8.84           246            9.38
No             Yes            No               85        3.24           331           12.62
No             No            .              2292       87.38          2623          100.00

Highlighted items from the recode output for skip patterns:

232 participants are coded as "1" (or “Yes”) for the new variable TREATOSTEO, which is correct based on OSQ060 and OSQ070.
Notice that 2,377 participants will be coded as “2” for the new variable TREATOSTEO -- 2,292 participants were not asked OSQ070 (i.e. OSQ070 is missing) because they answered “No” to OSQ060; 85 participants answered “Yes” to OSQ060 but “No” to OSQ070.
Based on their responses to OSQ060 and OSQ070, 14 participants s still have missing values for the TREATOSTEO variable.

Click here to close program output and highlights

Close Window to return to module page.