Multilevel Assessment of the Predictive Validity of Teacher Made Tests in the Zimbabwean Primary Education Sector

Abstract

The principal focus of this study was to undertake a multilevel assessment of the predictive validity of teacher made tests in the Zimbabwean primary education sector. A correlational research design was adopted for the study, mainly to allow for statistical treatment of data and subsequent classical hypotheses testing using the spearman’s rho. The variables that provided the bedrock for the study were aggregate test scores for pupils’ performance in:

1. Midyear tests and
2. End of year tests, in 2016.

The four null hypotheses that underpinned the study were tested on the basis of correlation coefficients computed using test scores generated through the aforementioned tests. After subjecting the hypotheses to the court of empirical evidence, through significance testing, some depictable results emerged. The major finding was that homogeneous results were observed, as all the four null hypotheses upon which the study was based were rejected. The major finding attested to the conclusion that teacher made tests in the Zimbabwean primary education sector were valid for prediction purposes. The study also observed that if tests being appraised were found to be valid, then the same tests were also reliable. In the light of the foregoing, recommendations on the broad applications of teacher made tests in educational programming were promulgated

Keywords: Predictive validity, Multilevel Assessment, Predictor, Predicand, Criterion, Null Hypothesis, Significance Testing, Educational Programming.