Diet History Questionnaire II and Canadian Diet History Questionnaire II: Coding Guidelines

Follow these guidelines when coding information in DHQ data files analyzed by Diet*Calc.

1. Formatted Questions instruct the respondent to select 1 choice from a list of possible answers. One character is used to code the response. Typically, this would be a digit, 0 to n-1, where n = the number of possible choices. If you have added questions that allow more than 10 responses then you must use letters rather than digits. Make this change by using "A" or "a" as the Start Code in Diet*Calc (General Formats on the Settings menu of Diet*Calc's dictionary editor). The number of valid responses and their meaning are defined in Formats. For formatted questions, data dictionaries and codebooks provided by the NCI use "M" to indicate a missing response and "E" for an error (multiple marks when only one mark is appropriate). Diet*Calc does permit other characters to be used for Missing or Error Codes. If a multi-oval question has a partial response, code the ovals as they were answered. For example, if the first 5 digits in the social security number are properly marked (e.g.,12345) but the last 4 are left blank, you should code the digits in the first 5 places and Ms in the last 4 (the field would be coded as "12345MMMM"). The following are exceptions to the coding rules for formatted questions:
   • "Are you male or female?" - any characters can be used to code the response to this question. Data dictionaries and codebooks provided by the NCI use "0" for male and "1" for female (to be consistent with other formatted questions since male is the first response listed on the NCI forms). However, the coding scheme for this question is defined separately to allow you to switch the order of the responses or use other characters (note: M and F can only be used if M is not used as the missing code). Edit the "Sex" variable in the data dictionary to modify the coding scheme for this question. The response to this question determines the sex-specific nutrient values used in the analysis. A default value is used when this question is skipped or not asked. This default value can be specified when editing the "Sex" variable or by selecting "Sex" from the dictionary editor's Settings Menu.
   • Dates - year is coded as printed on the questionnaire. For example, the year field in Today's Date has 5 choices and uses 4 character codes, "2007", "2008", etc., rather than "0", "1", and "2". The entire field should be filled with the missing or error character if applicable. For example, if M and E are used for missing and error then "MMMM" and "EEEE" should be used as appropriate. Months are coded with a 2 character code: 01, 02, 03,...,12, MM, EE (if M and E are the missing and error codes).
2. Filled in vs. left blank - in some cases the respondent is simply asked to mark an oval if it is appropriate. Leaving it blank is a valid answer (not a skipped question). For example, many questions on the DHQ ask the respondent to "mark as many as apply". For each oval, "1" is typically used to indicate "filled in" and "0" is used to code "left blank." However, alternative codes can be used (see General Formats on the Settings menu of Diet*Calc's dictionary editor).
3. Other Questions are fields not analyzed by Diet*Calc. These should be defined as "Other Question" variables in the QDD. No edit checks are performed on these fields, therefore any coding scheme can be used to code them.

Many fields in the Diet History Questionnaire (DHQ) use the same coding scheme or format. A format defines the number of choices for a question and the meaning of each choice. For example, the response choices for "How often did you eat..." questions are coded using one of several formats. The formats are set in the Questionnaire Data Dictionary (QDD). You may modify the existing formats using the dictionary editor in Diet*Calc.

• Frequency formats are used for questions that ask "How often did you eat/drink...."
• Size formats are used to code serving size questions, i.e., "When you ate (food), how much did you usually eat?"
• "Filled in" and "left blank" codes are used when the respondent is asked to mark an oval if appropriate (not filling in the oval is a valid response and is not a skip for this type of question). For example, some DHQ questions provide a list of choices and instruct the respondent to "mark as many as apply."
• Proportion formats are used to code questions that ask the respondent to specify how often (in fractions) the food was of a specific type. For example, the question "How often were your fruit drinks diet or sugar-free drinks?" has valid responses of "almost never or never", "about ¼ of the time", "about ½ of the time", "about ¾ of the time", and "almost always or always."
• Duration format is used in supplement questions to indicate length of time, for example, "For how many years have you taken multi-vitamins?"

A missing code indicates that the respondent skipped a question when a response was required. An error character indicates that the respondent marked two or more responses to a question where only one answer was appropriate. The following guidelines must be used for coding fields as missing or error.

1. Letters or symbols (such as '*', '#', or '!' ) must be used as the missing and error characters. If letters are used to code formatted responses, symbols must be used for missing and error. Missing and error characters may never be numeric.
2. When multiple characters are used to code a single oval, set all characters in the field to the missing character or to the error character when appropriate. For example, the year field in Today's Date uses one oval but is coded with four characters ("2002", "2003", etc.). If all choices are skipped, the entire field should be filled with the missing character ("....", for example).

You may not use the same character to represent both the missing and the error characters. In the DHQ II, '.' and '*' are the missing and error characters, respectively.