Comparison of Mixtured Factor Analysis and Factor Analysis
No Thumbnail Available
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Karışımlı faktör analizi homojen alt sınıflama yapmakla birlikte, faktör analizine göre iki önemli avantajı bulunmaktadır. Bunlardan ilki, her bir gözlemin hangi olasılıkla hangi alt sınıfa dahil olabileceğinin hesaplanmasıdır. İkinci önemli avantajı ise, her bir alt sınıf için parametre tahminlerinin elde edilmesidir. Bu çalışmada, evlilik sürelerine göre boşanmaların Türkiye'de iller bazında haritasının çıkarılması amaçlanmıştır. Bu nedenle, sınıflandırma amaçlı yaygın olarak kullanılan faktör analizi ile karışımlı faktör analizinin performansları karşılaştırılmıştır. Evlilik sürelerine göre Türkiye'deki boşanmalar; faktör analizine göre bağımsız 3 alt sınıfta, karışımlı faktör analizine göre ise bağımsız 4 alt sınıfta oluşmuştur. Faktör analizinde bazı illerin faktör yüklerinin alt sınıflardaki büyüklüklerinin bir birine yakın olması doğru sınıflandırma olasılığının düşük olduğunu göstermiştir. Bununla birlikte, karışımlı faktör analizinde, entropy doğru sınıflandırma olasılığı oldukça yüksek elde edilmiştir. Karışımlı faktör analizini faktör analizinden ayıran diğer önemli özellik, elde edilen her bir alt sınıf için ayrı parametre tahminlerini elde etmesidir. İllerin homojen alt sınıflara göre dağılımlarına bakıldığında, karışımlı faktör analizinin, faktör analizinde daha iyi sonuç verdiği saptanmıştır. Elde edilen bulgulara göre, karışımlı faktör analizinin, faktör analizine göre birçok avantajı saptanmıştır.
Although mixtured factor analysis makes homogeneous subclassification, it has two important advantages over factor analysis. The first of these is to calculate which subclass with which probability each observation might be involved. The second important advantage is that parameter estimates are obtained for each subclass. In this study, it is aimed at mapping out the divorce according to duration of marriage in terms of provinces in Turkey. For this reason, the performances of factor analysis, which is commonly used for classification purposes, and that of mixtured factor analysis were compared. Divorces in Turkey according to duration of marriage were formed in 3 independent subclasses according to factor analysis and in 4 independent subclasses according to mixed factor analysis. In factor analysis, the fact that the factor loads of some provinces are close to each other in the subclasses showed that the possibility of correct classification is low. However, in the mixtured factor analysis, the possibility of correct classification of entropy was found to be quite high. Another important feature that distinguishes mixtured factor analysis from factor analysis is that it obtains separate parameter estimates for each subclass obtained. Considering the distribution of the provinces according to homogeneous subclasses, it was found that mixtured factor analysis gave better results in factor analysis. According to the findings, many advantages of mixtured factor analysis compared to factor analysis were determined.
Although mixtured factor analysis makes homogeneous subclassification, it has two important advantages over factor analysis. The first of these is to calculate which subclass with which probability each observation might be involved. The second important advantage is that parameter estimates are obtained for each subclass. In this study, it is aimed at mapping out the divorce according to duration of marriage in terms of provinces in Turkey. For this reason, the performances of factor analysis, which is commonly used for classification purposes, and that of mixtured factor analysis were compared. Divorces in Turkey according to duration of marriage were formed in 3 independent subclasses according to factor analysis and in 4 independent subclasses according to mixed factor analysis. In factor analysis, the fact that the factor loads of some provinces are close to each other in the subclasses showed that the possibility of correct classification is low. However, in the mixtured factor analysis, the possibility of correct classification of entropy was found to be quite high. Another important feature that distinguishes mixtured factor analysis from factor analysis is that it obtains separate parameter estimates for each subclass obtained. Considering the distribution of the provinces according to homogeneous subclasses, it was found that mixtured factor analysis gave better results in factor analysis. According to the findings, many advantages of mixtured factor analysis compared to factor analysis were determined.
Description
Keywords
İstatistik, Statistics
Turkish CoHE Thesis Center URL
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
71