MAHDALENA, 090013954M (2002) KOMPARASI ANALISIS DISKRIMINAN PADA DATA ORDINAL METODE RANK DENGAN METODE DISKRIT (APLIKASI PADA DATA STATUS IMUNISASI BADUTA DI KABUPATEN TAPIN DAN HULU SUNGAI TENGAH KALIMANTAN SELATAN). Thesis thesis, UNIVERSITAS AIRLANGGA.

Preview	Text (ABSTRAK) gdlhub-gdl-s2-2006-mahdalena-714-tkm_09-03.pdf Download (530kB) | Preview
Preview	Text (Fulltext) 35791.pdf Download (1MB) | Preview

Abstract

Complex problem requires the scientists to investigate them by involving several relevant variables. Considerable variables in the research make univariate analysis method can not do any analysis. Therefore, multivariate analysis method is apparently needed, one of which is discriminant analysis. However, in fact, assumptions which have been determined can not be always satisfied, particularly in ordinal data. Thus, such a problem can be solved without requirements (assumptions). This is a ststistical study for comparing results of discriminant analysis in ordinal data and two approaches, noticably wilcoxon rank and discrete methods as criteria of minimum classification errors. Discriminat analysis composed of wilcoxon rank and discrete methods had some advantages and disadvantages. From wilcoxon rank method, discrete variables which discriminated complete and incomplete immunization status groups could be known, namely variable X3 = mother's knowledge level of children under two years on immunization and variable X6 = attitude/perception toward immunization. Since both variables had value of Ft > 3.84 that is, in the first stage, Ft for X, = 32.429, while int the second stage, Ft for X3 = 22.514 and X6 = 6.505. discriminant function errors at 132 respondents (37%), suggesting that discriminant function could serve as discriminant among research categories, In discrete method, discrete variables could not be known but it only had a combination of same varying discrete variables that can be seen from largest discriminant score. However, the discrete method had total classification errors at 113 respondents (31,65%). This means that disciminant function was more relevant as the discriminant among the research categories. When comparison between wilcoxon rank and discrete method is made, then the discrete method is preferred in the grouping problem since it has smaller value in classification errors than that of the rank method.

Actions (login required)

View Item