Reliability Vs. Validity: Difference & Types

In a research design, especially in a quantitative research, reliability and validity are highly important. These two concepts are very closely related, although their meanings are different. The differences between the two are very subtle. In a research, a measurement can be reliable but not necessarily be valid, however, if a measurement is valid, then it is considered to be reliable.

Reliability

Reliability of a technique, method, tool or research instrument implies how consistent it measures something. If by using the same technique or methods, same outcome is consistently achieved under similar circumstances, then the measurement is said to be reliable. For example, if a thermometer displays same temperature of a same liquid sample under identical conditions, then the results can be considered as reliable. Thus, it can be said that reliability refers to the measure of consistency or stability of the test scores.

Different statistical tools are used to measure reliability, such as, the Kuder-Richardson 20 and Cronbach’s alpha. The Kuder-Richardson 20 measures the internal reliability for binary tests and Cronbach’s alpha measures the internal reliability for the tests having multiple answers.

Types of reliability

Reliability is categorised as internal and external reliability. Internal reliability of a test measures how well the test actually measures what it is supposed to measure. On the other hand, external reliability of a test implies how well the test can be generalized beyond what it is meant for.

The other types include test-retest and interrater. The test-retest reliability represents the consistency of a test measure across time and interrater is the reliability which represents the consistency of the measure across observers or raters.

Reliability coefficients

A reliability coefficient presents the measure of how well a test or a research instrument measures the achievement. The proportion of variance in the observed scores which is attributable to the actual scores is shown in the reliability coefficient.

• Cronbach’s alpha is the most common measure of internal reliability.
• Simple correlation between two scores obtained from same individual is also a reliability coefficient. This represents the test-retest reliability if the tests are conducted at different times. If different types of tests are conducted on the same day, that can give parallel forms reliability.
• Pearson correlation is the measure for estimating theoretical reliability coefficient between parallel tests.
• Spearman Brown formula is used for measuring reliability for split-half tests
• Cohen’s Kappa is used for measuring interrater reliability.

The range of the reliability coefficient lies between 0 and 1. In maximum cases, the score of 0.7 is considered as acceptable. A score higher than 0.7 is taken to be good, and a score less than 0.7 is not a good score for reliability.

Validity

Validity refers to the incidence that how well a test or a research instrument is measuring what it is supposed to measure. A research having high validity implies it is producing the results corresponding to real properties, variations and characteristics of the different situations. Validity is more difficult to assess than reliability, however, it can be assessed by comparing the outcomes to other relevant theory or information.

Types of validity

There are many types of validity in a research study. However, three major types of validity are construct, content and criterion. Some other types of validity are: Composite, Concurrent, Convergent, Consequential, Curricular and Instructional, Ecological, External, Face, Formative validity & Summative Validity, Incremental Validity, Internal, Predictive, Sampling, and Statistical Conclusion Validity.

• Construct validity: In this type of validity, the adherence of a measure to some existing knowledge and theory of the research concept is measured. For example, a survey questionnaire on assessing self-esteem of the participants can be examined by measuring other known traits or assumed to be associated with the concept of self-esteem, like, optimism and social skills. If there is high and strong correlation between the scores of self-esteem and other traits, then that would imply a high construct validity.

• Content validity:This validity examines the extent to which the tool, method or measurement covers all the aspects of the concept being measured. For example, if in an examination for measuring the proficiency of the students on French language, only reading, writing and speaking are measured and listening is not measured, then the test is considered to be having low content validity as listening is also a part of language proficiency.

• Criterion validity:In this validity, the extent to which the outcome of a specific measure or tool corresponds to the outcomes of other valid measures of the same concept is examined. For example, a survey is being conducted by a news agency for assessing the political opinion of the voters in a town. If the survey results are accurate in predicting the actual outcome of the election in that town, then it is considered that the survey has a high level of criterion validity.
• The validity of a causal relationship, that is, cause and effect relationship, the internal and external validity are measured.

Thus, it is evident that these two concepts are highly important for a research paper study to assess whether the outcomes are consistent and whether the measurement instrument is precise to what it is meant to measure.