Kelly's Non-Parametric Factor Analysis
Bibliographic Data
| ID | 13658352 |
|---|---|
| Authors | A Z ARTHUR (University of Canterbury, corresponding author) |
| Year | 1965 |
| Volume | 16 |
| Issue | 3 |
| Pages | 922-922 |
| Publication date | 1965-06-01 |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Psychological Reports (JOURNAL) |
| Journal identifiers | ISSN: 0033-2941 • E-ISSN: 1558-691X |
| Publisher | SAGE Publishing (PUBLISHER • US) |
| DOI | 10.2466/pr0.1965.16.3.922 |
| PMID | 14303031 |
| OpenAlex | W2339805941 |
| Language | EN |
| Citations received | 1 |
Kelly ( 1955) has proposed a time-saving method of non-parametric factor analysis for use with his Rep Test, a measure of personal constructs. The analysis appears to be most closely related to the major-factors method of Thurstone. The present study was undertaken to check the extent to which the non-parametric technique produces a similar factor structure to the parametric method of Thurstone's centroid procedure. Method.-10 adults attending an evening class were each given the Rep Test. A modified form suitable for self-administration was ~lsed and Ss were required to produce 20 figures and to apply to them 20 constructs. From each S's grid two factors were extracted, using Kelly's method and again using Thurstone's method. Centroid factor plots indicated that rotation was possible in 4 cases to increase the similarity of both factors obtained by the 2 methods. This was done by the graphic method. As indices of similarity the loadings obtained by the two methods from each S were compared by means of Spearman's rhos. Results.-As regards the similarity of the first two factors, out of 10 correlations 4 are above .80 and only one is below .50. These results may be considered as satisfactory indications of similarity between the two methods. As regards the similaricy of the second faccors, only 2 correlations are above .40 and 5 are below .20. There is thus little agreement between the second factors. It is to be noted that in three instances Kelly's method produced a second factor which was identical with the first. This is probably due to the fact that the residuals for computing the second factor are arrived at by deducting only those variances which are at an arbitrary level of significance. This aspect of the non-parametric analysis therefore needs improvement
Factor (programming language · Statistics · Behavioral and Psychological Studies · Cognitive and psychological constructs research · Computer Science · Mathematics · Psychology · Sensory Analysis and Statistical Methods
| Citations per year | 0.02 |
|---|---|
| Citation span | 1974 - 1974 (1) |
| Highly cited | No |