Harokopio University
Διερεύνηση του ρόλου των ορθογωνίων και μη ορθογωνίων, μετασχηματισμών των αξόνων στην παραγοντική ανάλυση, σε σχέση με την επαναληψιμότητα των εξαγόμενων προτύπων: εφαρμογές σε ιατροβιολογικά και διατροφικά δεδομένα
Abstract
dc:descriptionINTRODUCTION: Factor analysis is a technique of reducing sample variables that creates a new set of variables, the common factors interpreting the sample. The starting axes are then rotated until some other suitable position is reached, in order to make the common factors interpretable. Before applying these results, the common factors should be checked for their accuracy. Repeatability (or reproducibility) is the most important parameter in determining the reliability of a method. It refers to how close the values are to each other from a series of measurements made exactly the same way. However, the degree of generalization of the extracted patterns to the population has rarely been studied. ΑΙΜ: The aim of the proposed doctoral thesis is to investigate the role of orthogonal and non-orthogonal rotation of the axes on the factor analysis, in regard to the repeatability (or reproducibility) of the extracted factors. ORIGINALITY: Although in the international literature the use of factor analysis is extensive nowadays, the degree of repeatability of the extracted patterns (factors) and the role of rotation (i.e., orthogonal or non-orthogonal) has not been investigated on a theoretical level. METHODS: Simulation models will be used in R, in which we will run 1000 times to build confidence intervals. After applying factor analysis and extracting the common factors we will add an error. The basic approach is to check the stability of the factors, which depends on the sample size and the number of cases per variable. We will test the power of the results by re-applying the method and comparing the results. The application will be based on epidemiological data. RESULTS & CONCLUSIONS: From the simulation studies it was observed that applying non-orthogonal rotation and, specifically, Promax rotation gives more robust results, comparing them to those with no rotation, and specifically if a large number of extracted factors were selected (more than 30). In most cases the results were very similar to the orthogonal rotation methods (Varimax and Quartimax). It is also noteworthy that when no type of rotation was applied there was much less repeatability (or reproducibility), that is, the results could not be generalized in most cases.
Degree
thesis:*- Grantor dc:publisher
- Harokopio University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Panaretos, Dimitris
- Πανάρετος, Δημήτριος
Subjects
dc:subject × 20- Παραγοντική ανάλυση
- Πολυμεταβλητή ανάλυση
- Μέθοδοι περιστροφής
- Επαναληψιμότητα
- Αναπαραγωγιμότητα
- Προσομοιώσεις
- Factor analysis
- Multivariate analysis
- Rotation methods
- Repeatability
- Reproducibility
- Simulations
- Φυσικές Επιστήμες
- Μαθηματικά
- Ιατρική και Επιστήμες Υγείας
- Επιστήμες Υγείας
- Natural Sciences
- Mathematics
- Medical and Health Sciences
- Health Sciences
Rights
dc:rights- Statement dc:rights
-
- BY_NC
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/47271