Pulos, Steven M.
McDevitt, Teresa M.
University of Northern Colorado
Type of Resources
Place of Publication
University of Northern Colorado
How parents interact with their children impacts many crucial facets of children‟s lives. Over the last 4 decades, researchers have identified four different parenting styles: authoritative, authoritarian, permissive, and disengaged. Hundreds of studies conducted all over the world, have identified correlations between parenting style and many different child outcomes (e.g., academic achievement, coping, self-esteem, and decision-making abilities). The interpretability of these results has been limited by the unknown psychometric properties of the instruments used. The building block of any psychometrically sound measure is instrument reliability, which is the degree of non-error variance in the instrument‟s results. Sound estimates of reliability may be obtained from the synthesis of existing research. To this end, Reliability generalization (RG) focuses on the estimate of reliability reported by primary studies to produce a measure of central tendency of the reliability of the data collected by the instrument across different studies and the impact of identified moderator variables on the magnitude and distribution of reliability estimates. This reliability generalization dissected the most commonly used measure to identify parenting style--the Parental Authority Questionnaire (PAQ; Buri, 1991). Two different statistical methods for determining the reliability of the PAQ were compared: the random-effect model and the varying coefficient model. In this investigation, all six scales of the PAQ were found to be reliable at a level generally acceptable for research purposes, but not necessarily reliable to the degree necessary when making decisions that affect individuals. Moreover, each of the six scales was not equally reliable. Seven moderator variables were identified that significantly affected scale reliability; (a) scale mean, (b) scale variance, (c) percent of male participants, (d) developmental level of participants, (e) the language the PAQ was administered in, (f) the type of publication, and (g) the year of publication. A comparison of the two different RG meta-analytic methods revealed that both models yielded similar results with regard to scale reliabilities. However, the varying-coefficient model provided smaller confidence intervals and was more sensitive at detecting moderator variables. This dissertation has implications for both those who intend to use the PAQ as well as those who conduct reliability generalizations. First, before researchers plan interventions based on the results of studies that seem to correlate parental authority with various other psychological, sociological, or educational constructs, care must be taken to ensure that sound methodological practices are in place throughout the research process. These practices should include calculating reliability each time an instrument such as the Parental Authority Questionnaire is administered. Second, this research yielded valuable information for meta-analytic researchers by demonstrating, with actual study data, how methods of analysis in reliability generalization could differ in not only identifying reliability estimates but variables that moderate reliability as well.
Copyright belongs to the author.