Factor analysis in constructing and validating personality tests

Reliability & Validity

For the lay person, the notion of personality is often derived from components of an individual's character or make up that has the ability to elicit positive or negative reactions from other individuals. The person who has a propensity for positive reactions from others is often thought to have a 'good' personality. Conversely, the person who tends to elicit not so favorable reactions from others may be thought to have a 'bad' personality. However, when behavioral and social scientists seek to describe and define personality, the terminology used is far more rigorous that describing simple social skills (Cohen, Montague, Nathanson & Swerdlik, 1988). As such, constructs such as personality traits, personality states and personality types have been studied as a means of provided clinically accurate ways in which to define personality.

Nevertheless, there is no one globally accepted definition of personality within the scholarly literature. McClelland (1951, p. 69) defined personality as "the most adequate conceptualization of a person's behavior in all its detail." While Menninger (1953, p. 23) defined personality as "the individual as a whole, his height and weight and love and hates and blood pressure and reflexes; his smiles and hopes and bowed legs and enlarged tonsils. It means all that anyone is and that he is trying to become." Although no one definition of personality has been globally accepted in the scholarly world, there are some components and constructs of personality that have been more widely accepted.

Factor Analysis in constructing Personality Testing

Because there are so many ways in which to describe an individual's personality, those interested in personality frequently use a statistical tool to simplify the enormous amounts of information available by placing similar information into clusters known as factor analysis. The premise behind factor analysis suggests that if two or more characteristics correlate, they may reflect an underlying trait that is shared; thereby creating patterns of correlations that reveal the trait dimensions existing beneath the measure qualities (Tabachnik & Fidell, 2005). Factor analysis is a more complex version of correlation, however, in the sense that instead of examining correlation between just a few variables, factory analysis uses a great number of correlations among a great number of variables (Kline, 1994).

In order for factor analysis to be completed, the researcher first collects data on many variables across a significant number of individuals. The data can be collected in any number of ways, but what is important is that the same data is collected from everyone participating. Upon collection of the data, the researcher then calculates the correlations between every conceivable and possible pair of variables. In this way, the factor, in personality research is commonly viewed as a reflection of a personality trait (Gorusch, 1983). Researchers use factor analysis to construct and refine personality tests. Although the label of a factor is primarily something that has been inferred from a cluster of correlating variables, there is the assumption that personality tests scores directly reflect the individual's personality traits with little to no error. Factor analysis is determined useful in personality testing because it simplifies the various ways a person is understood by reducing the information into smaller more manageable sets of personality traits. Factor analysis provides a basis or contextual framework that perhaps some traits are more important than others when derived from large highly correlating clusters. And factor analysis is very useful in creating personality measures. However, it is important to remember that factor analysis' usefulness is contingent upon the information that the researcher inputs; resultantly, the facts that emerge are largely dependent on the kind of data collected or the variables that were included in the process of analysis (Kline, 1994).

Reliability of Personality Tests

Reliability in personality testing is a measure of consistency. If a measure (trait) on a personality test was considered reliable, then there is the expectation that almost identical scores would be achieved on the retest. The smaller the variance between the two scores, the more accurate or reliable the measure is said to be. Reliability of a measure is determined on the correlation coefficient which has a range from +1.00 to -1.00. The correlation coefficient measure the strength between the two variables. For example, if a coefficient approaches plus or minus 1.00, then a strong relationship is determined with a +1.00 reflecting a positive relationship and a -1.00 representing a negative relationship. If the result would be 0.00, then no relationship is indicated (Joint Committee, 1999).

The most frequently used method of establishing reliability in personality testing is the test-retest method. In order to achieve this, the same individual is tested at two separate points in time and a correlation coefficient is established to ascertain if the scores on the initial test are related to the scores on the subsequent test. A high correlation coefficient indicates to the researcher that the individual scores on the initial test are very similar to the scores on the subsequent test. The importance and significance of reliability, then, demonstrates that if the measurements are not consistent, there is not value or benefit to be derived from the personality test (Joint Committee, 1999).

Applicability of Personality Tests

With regard to whether or not personality tests can be used as a means of projecting potential or continued employment success, it is first most important to ensure that the personality test utilized has both a fairly high level of validity and reliability. Without these two factors, the usefulness of the test results will be significantly diminished because of the low consistency in the results and the test failing to measure what it is intended to measure. The data collected from these personality tests, then, would be of little use from a historical perspective for continued employment. If personality tests are used as a pre-employment measure in the employment of person that requires certain qualities or behavioral traits, again the tests should be reliable and valid. If an instrument is unreliable, it would then prove difficult to relate the behaviors to any kind of theoretic behavioral model as the rules will vary instead of being consistent.

Although personality tests cannot determine with 100% accuracy how an individual will or continue to perform tasks, it does provide some valuable information as to what may be expected regarding performance, particularly if there is historical tests data to compare it to. However, because of the level of predictability that can be derived using personality tests that are determined reliable and valid, a certain amount of conjecture can be used in assessing an employer's predicted performance.

Conclusion

Personality is a construct that denotes a variety of meanings and reflects a number of characteristics. Although no one global definition of personality is universally accepted, there are ways in which personality can be assessed to derive information about an individual's qualities that have proven particularly assistive in employment. However, in order for personality testing to be effectively utilized, the measure must be both reliable and valid. Subsequent to the establishment of these two very important test characteristics, employers are then able to operate with some degree of confidence in using personality tests as a measure of pre-employment and continued employment predictability. Although personality testing should never be the sole method by which hiring or maintaining an employee are based on, valid and reliable personality assessments provide invaluable information that can be used in the employment process.