Reliability and validity are two important characteristics of any measurement procedure.
Reliability has been defined as ‘the extent to which results are consistent over time… and if the results of a study can be reproduced under a similar methodology, then the research instrument is considered to be reliable.’ (Joppe 2000). This means a test is considered reliable if the same results are produced repeatedly, if it were to be carried out again. The more consistent the results produced, the higher the reliability of the measurement procedure.
Reliability is addressed in a variety of ways. These include:
- Inter-Rater Reliability – is the variation in measurements taken by different people using same methods. In order to ensure reliability, the degree of variation must be small.
- Test-Retest Reliability – is established by comparing scores of the same individual, to calculate a correlation. There must be a strong correlation to ensure reliability.
- Split-Half Reliability – is obtained by dividing up the test into two comparable halves, in order to calculate consistency between the two scores. Consistency must be present in order for the test to be considered reliable.
Despite having methods to ensure reliability, there are issues which arise which can affect the reliability of the test and results. The researchers are human, and that means the experiment is open to human judgement and error. However, this can be solved by carefully reporting methodology in the study and, if using qualitative methods, double coding.
These solutions make reliability much easier to assess.
Validity, on the other hand, determines whether the research truly measures what it intended to measure, (Joppe, 2000). This means it looks at the extent to which a test measures what it claims to measure, and therefore, answers the research question or hypothesis. How valid a test is, depends on the purpose of the research.
Validity is addressed in a variety of ways, and include:
- Content (face) Validity – is the degree to which a test measures an intended content area. It must measure what it claims to measure, in order to be considered valid.
- Concurrent Validity- measures the extent to which a correlation exists between a new measure and a standard measurement procedure. The scores should be directly related in order to obtain validity.
- Convergent Validity – is the degree to which scores obtained from two different methods of measures. There must be a strong relationship, for validity to be demonstrated.
Despite having methods in place to ensure validity, there are threats. There are two main threats: experimenter bias and demand characteristics. Again, the researcher is human and this means the study will always be open to human error. The experimenter may influence the outcome of the research because of his/hers expectations regarding the results. This however, can be solved through the use of single blind or double blind studies, in which the researcher has no idea what the predicted outcome is. Again, the participants are human, and this can lead to the problem of demand characteristics, in which participants behave in a different way. Participants normally modify their behaviour in response to the fact they are participating in a study and are aware they are being measured. They strive to be a good participant. Although it is essentially impossible to prevent participants from modifying their behaviour, there are methods in place which can reduce this effect. Solutions include: using observations or concealing the measurement procedure.
Despite being very different, both reliability and validity are important in research. As the saying goes ‘a valid test is always reliable but a reliable test is not necessarily valid’, but it is important to ensure that both reliability and validity are demonstrated.