Questionnaires are most widely used tools in specially social science research. Most questionnaire’s objective in research is to obtain relevant information in most reliable and valid manner. Therefore the validation of questionnaire forms an important aspect of research methodology and the validity of the outcomes. Often a researcher is confused with the objective of validating a questionnaire and tends to find a link between the reliability of a questionnaire with the validity of it.
The reality is that reliability and validity are two different aspects of an acceptable research questionnaire. It is important for a researcher to understand the differences between these two aspects. In its simple explanation, reliability of a questionnaire seems to emerge from the quality of the questionnaire. On the other hand validity seems to emerge from the internal and external consistency and relevance of the questionnaire. In other words reliability of a questionnaire refers to the quality of tool (read questionnaire) while validity refers to the process used to employ the tool in use, i.e. the process used to conduct the questionnaire. There are several dimensions to the process of employment of a questionnaire in use. Some of the important dimensions are discussed in the following paragraphs.
A major aspect of validation of a questionnaire refers to common validity of the questionnaire. The most common elements widely used in questionnaire validation are –
Known Group Validity – refers to the extent to which an instrument can demonstrate variability of scores which vary on a certain known variables.
Construct Validity – refers to the extent to which an instrument can demonstrate the measure of the intended construct.
Content Validity – refer to the extent to which an instrument covers all aspect of social problem under study
Criterion Validity – refers to consistency with the gold standard questionnaire
Variables may have correlation but this correlation should be optimal. Most commonly correlation tests are aimed at finding interclass correlation, between group correlations. Correlation mainly provides measure of internal consistency for validating the questionnaires. Some of the common correlation test for validating questionnaire relate to following
Inter class correlation coefficient – It refers to the ratio between interclass variance to total variance.
Cronbach Alpha – Is the measure of the correlation between items of the test. It is the homogeneity of the test. Experts agree that items in a test are moderately correlated. This way these are expected to measure all aspects of a single trait being tested. If the correlation is too low it may indicate that items refers to not one trait but two or more different trait. On the other hand a very high correlation refers to one of the items being redundant for the test
Discriminant correlation – refers to the extent to which a measure of a research attribute is related to measure of a different attribute which is not intended to be measured.
Bias is more problematic than random error and can be intentional or unintentional. The bias is related to characteristics of investigator, observer or instrument. Unintentional bias is of bigger concern. It is to be avoided by uncovering its source and re looking at design instrument or using a method to avoid it.
A validated questionnaire is one that has undergone validation procedure to show that it accurately measures what its objective is, regardless of the respondent’s status, timing of response, different investigators. The instrument is compared with the Gold Standard, if available. It is also compared with other sources of data. The reliability is also tested. Even if the questionnaire is not fully valid (which is rare), reliability of the questionnaire has its own value. If the reliability is there it offers an opportunity to compare results with other studies.