Write 5 pages thesis on the topic tests of reliability and validity. It is used to measure just one single phenomenon Items are said to be internally consistent if these items are highly correlated (DeVellis, 2003). Coefficients of internal consistency deal with variance as a result of variation in item content as a primary source of error variance. At the same time, coefficients “show how much the variance is systematically based on a common concept measured by the test as a whole” (Guion and Highhouse, 2006, p. 129). The main strength of internal consistency validity is its applicability in education when teachers can not use alternate forms of the same test and its convenience since only one administration and one test form is required (Oermann and Gaberson, 2006. Guion and Highhouse, 2006). The main weakness of internal consistency is associated with the complex field of human temperament and dynamics where high internal consistency has been found to be an antithesis to high validity, especially when the variables being measured are broad (Kline, 1986). Another weakness is that even if it is used to measure only one variable, estimators of internal consistency such as Cronbach’s alpha are functions of two parameters (Robins, Fraley and Krueger, 2007). It was also revealed in a research study by Puhan, Bryant, Guyatt, Heels-Ansdell, and Sch?nemann (2005) that internal consistency is a poor predictor of responsiveness.
Split-half. The test for split-half reliability measures the internal consistency of items with a high degree of fluctuation. As the name implies, the items are divided in half, usually on an odd-even basis, and the two halves are correlated. The main strengths of split-half reliability are the impossibility of the effects of time to intervene in the results since the test is administered only once. and that the respondents are exposed to the same physical, mental, and environmental influences. Meanwhile, there are also a few limitations : (1) the fewer items resulting from the splitting into halves results in a wider dispersion of the responses. (2) as a result of the variability, the responses do not easily stabilize around the mean. (3) the need to determine if the two halves of the test actually measure the same thing (Groth-Marnat, 2003. Portney and Watkins, 2009).