3.5 Missing Types and their Meanings

  • Print
    • In TwinLife, missing values are delivered in a differentiated way. Table 3 gives an overview of the standard missing codes and a short explanation of their meaning.
    Value
    Last Name
    Explanation
    -99:
    not specified (refused to answer)
    The respondent explicitly has refused to answer the question.
    -98:
    don‘t know
    The respondent has explicitly stated not to know what to answer.
    -96:
    mixed missing values – e.g. don‘t know/not specified
    Here, the instrument has not differentiated between two or more types of missings. This is mostly the case in the paper and pencil questionnaires of the first survey wave.
    -95:
    doesn't apply (screened out)
    It indicates that the question was not intended for the respondent based on the filter conditions, for instance,. due to the respondent’s age.
    -94:
    technical error / faulty insertion
    In most of these cases, the filter condition of the question was programmed incorrectly so that respondents have falsely received or not received a certain question.
    -93:
    unclear classification of system missing (only for paper-and-pencil questionnaires)
    Here, it is not possible to determine what the reason for a missing value is (whether it does not apply, the respondent does not know or whether they do not want to reply).
    -92:
    no participation in survey module
    The respondent has not participated in the survey module / questionnaire, either because the questionnaire did not apply for the respondent or they refused to participate.
    -90:
    no participation in survey wave
    The person has participated in past data collections or may participate in future data collections, but not in the current one (all variables have the missing -90 in this wave for this person).
    -87:
    multiple answers
    The person has given multiple answers in questions where only one answer is possible (usually in paper and pencil questionnaires).
    other -80s:
    … have different meanings and can contain valid information (see below)

    Table 3. Missing codes.

    Please note that the missing codes -80 to -89 (except for -87) have special meanings that can differ between variables and can be of importance for analyses. Generally, they contain ‘valid’ information but the answers are not direct answers to the question and are therefore coded as negative missings.
    One example is the variable emp0800 in wid == 3 (data collection F2F 2) “When did you leave your last job?” – The answer that the interviewee has not been employed until now is not a direct answer to the question but contains valid information and is therefore coded as a negative missing (-81).
    In contrast, the answer “I already have children” on the question “Having children: How likely do you think you are to achieve this goal in your lifetime?” (lgd0204) in wid == 3 (data collection F2F 2) is a valid and analyzable answer to the question and is therefore coded with a positive code outside the actual answering scale (96).
    You should check all variables for positive and negative meaningful missing values before using them for analyses.