A content analysis of selected multivariate statistical textbooks

The purpose of this research was to use content analysis as a means of analyzing multivariate textbooks. It was demonstrated that content analysis can be used to numerically analyze multivariate statistical textbooks. The sample for the analysis consisted of three present day multivariate statistical textbooks. The textbooks studied were for use by graduate social science students. Only textbooks published in the United States were used. The textbooks were only to have a mathematical level of high school algebra. The original listing of textbooks came from publishers of mathematical and statistical textbooks. In addition, textbooks listed on amazon.com were also considered. Through a winnowing process using publisher reviews, text table of contents, reader's summaries and reviews, and inspection, three books were chosen that met the criteria.; A content analysis software program was used for the actual analysis. Multivariate statistical techniques chapters investigated were canonical correlation, discriminant analysis, factor analysis, MANOVA, and multiple regression. Key statistical words were determined by the number of times they appeared in the text coupled with their statistical significance. Sentence counts were obtained for one to five combinations of keywords. These results compared and contrasted these combinations in relation to the five statistical techniques and the texts examined. Also studied was the use of mathematical emphasis and power in the texts. Conditional words, contrasting connective words, connectives adding idea words, result connective words, and cause or reason words were also included in this research. In addition negation words with up to four word statistical keyword combinations were examined. Lastly, concatenated negation sentences were studied. This was accomplished by examining the negation sentences to determine their sentence type or category. A coding guide was constructed to help determine sentence category by observing key words or phases in the sentences. Two coders determined a 72.1% reliability of the concatenated negation sentences. There were differences per textbook and statistical technique among the one, two, three, and four concatenated negation sentences using Spearman rho. These results then demonstrate that content analysis is capable of being used as an evaluation tool to analyze multivariate statistical textbooks.