Validity means accurate or error free conclusion(s) from the data. Technically, we can say that a measure leads to valid conclusion from a sample that can be taken as valid inference about the population. When we talk about validity, we are talking about four major types:
1. Internal validity
2. External validity
3. Statistical conclusion validity
4. Construct validity
Internal validity: When the relationship between the variable is casual, than it is called internal validity. Internal validity refers to the casual relationship between the dependent and the independent variable. In internal validity, we are concerned with the factor responsible for change in the dependent variable. It is related to the design of the experiment, such as when it is used for the random assignment of treatments.
External validity: External validity is when there is a casual relationship between the cause and effect that can be generalized or transferred to different people, to different treatment variables and to different measurement variables.
Statistical conclusion validity: Statistical conclusion validity occurs when we talk about the inference about the degree of the relationship between the two variables. For example, it is used when two variables are studied and we want to draw a conclusion about the strength of the relationship between the variables. When we arrive at the correct decision about the strength of the relationship for both of the variables, then it is said to be statistical conclusion validity. Statistical conclusion validity has two major types of errors:
Type one error: Type one error occurs when we accept the hypothesis, but that hypothesis is inaccurate. It also occurs when we say that there is a relationship between the two variables but in reality there is no relationship between them.
Type two error: Type two error occurs when we reject the hypothesis that is true or when there is no relationship between variables, yet we say that the relationship exists.
Power analysis is used to detect the relationship in statistical conclusion validity. When we are using statistical conclusion validity, we come across several problems. One of these problems occurs when we use a small sample size. In a small sample size, there is a possibility that the result will not be accurate. To overcome this problem, the sample size should be increased. Violation of the statistical assumption is also a threat for statistical validity. If we use a biased value in analysis, then the results may not be accurate. If the wrong statistical test is applied, then the conclusion may not be accurate.
Construct validity: Construct validity is when the construct is involved in predicating the relationship for the dependent variable. For example, in structural equation modeling, when we draw the construct, then we assume that the factor loading for the construct should be greater than .7. Cronbach's alpha is used to draw the construct validity. .60 is considered acceptable for exploratory purposes, .70 is considered adequate for confirmatory purposes, and .80 is considered good for confirmatory purposes. If the construct satisfies the above assumption, then the construct will contribute in predicting the relationship for dependent variables. Convergent/divergent validation and factor analysis is also used to test construct validity.
Relationship between reliability and validity: A test that is unreliable cannot be valid and a test that is valid must be reliable. Reliability is necessary but not a sufficient condition for validity. Thus, validity plays a major role in analysis and in making accurate decisions.
The following are overall validity threats:
1. Insufficient data collected to make valid conclusions
2. Measurement done with too few measurement variables
3. Too much variation in data or outlier in data
4. Wrong selection of sample
5. Inaccurate measurement method taken for analysis