Taylor’s study (see above) is an empirical study of interest to the education profession in general. The areas of specialization include the affective domain and home schooled children.
Purpose. “This study sought to analyze the relationship which exists between home schooling and the self-concept of children in grades four through twelve” (p. 5).
Instrument. The Piers-Harris Self-Concept Scale(PHSCS) “…is an eighty-item checklist questionnaire in which the respondent answers simple declarative statements with a dichotomous rating of Yes or No, thus indicating whether or not a specific stimulus applies to him/her” (p. 111). Taylor dedicated much space in his dissertation to examining various self-concept instruments and explaining why he chose the PHSCS.
In terms of reliability, the PHSCS was reported to have an internal consistency of .90. The median test-retest reliability was .73. The author pointed out that the six subscale reliability values were considerably lower than for the global scale.
Taylor referred to a study which gave values of .56 and .61 for the predictive validity of the PHSCS. He also mentioned that perhaps the idea of criterion validity is not applicable to the measurement of self-concept. Several studies say that the PHSCS is a valid measure of the central core of personality. The convergent validity between the PHSCS and four other related instruments included values ranging from .60 to .71.
Statistics Used. For hypotheses (1-3) involving population norms, t tests were used. Multiple linear regression, stepwise, best subsets, and backwards methods, was used for the others (4-16). The assumptions of multiple regression were not discussed by the researcher.
Sampling. This author assumes that the general population was fourth through twelfth grade students in the U.S. The group used to norm the PHSCS was 1,183 public school students in Pennsylvania. The home schooled group came from an original list of 45,000 supplied by Holt Associates, Inc. and Hewitt Research Foundation. From this list, Taylor randomly selected 2,000 families. Of these, he estimated that about 500 “…might have qualified as participants” (p. 128). With a return rate of about 45%, there were finally 224 home schooled individuals included in the study. These subjects evenly represented the four geographical regions of the U.S., the researcher reported.
The sample size of 224 was sufficient for t tests. A rule of thumb for multiple regression is 14 subjects for the first variable and 10 for each variable thereafter (Wayne Courtney, personal communication, June 24, 1986, Oregon State University). Thus, with 14 variables in this study, the minimum sample size should have been 144. The sample was adequate.
General Comments. Taylor’s purpose was clear, a convincing theoretical framework for the study was established, the selected instrument was appropriate, the sampling matrix was clearly described, and the sample size was adequate for the statistics used. The statistics used appear to be appropriate, but the assumptions of regression were not addressed. The results were clearly explained and the conclusions generally followed from the data in a logical fashion. Taylor’s recommendations, including a similar study of achievement levels and longitudinal studies of the self-concept of home schooled children, are worthy of consideration.