Biased Items in Admission Test of Faculty of Medicine, Prince of Songkla University

Abstract

The purposes of this research were to study the bias and the consistency of
detecting item bias in admission test of Faculty of Medicine, Prince of Songkla
University by a direct system for students of the 14 southern provinces . The sampling
groups were categorized into types of schools dividing into students of General schools
and those of Private Islamic schools, and into sites of schools dividing into students
whose schools locate in 11 southern provinces , and those whose schools locate in 3
southernmost provinces. The samples were 410 students in Admission Test of Faculty
of Medicine, Prince of Songkla University by a direct system for students of the 14
southern provinces during the academic year 2009. They were from the estimating the
sample by using minimum examinees according to Krejcie and Morgan’s table on a 5
percent level of acceptable error , and a 95 percent confident level. The research tool
for data collection was Admission tests of Faculty of Medicine, Prince of Songkla
University by a 14 southern province direct university admission system during
academic year 2009. The tests consisted of 7 subjects , each with 2 issues; totally 14
issues. They were Mathematics , English, Thai, Social Studies ,Physics, Chemistry, and
Biology, and were analyzed by using the classical test theory with Analysis of Variance
( ANOVA ) and Chi -Square methods.
It was found that Mathematics, there were 2 issues, each with 40 items.
The first issue analyzed by Analysis of Variance had 10 biased items on types of schools
and 7 biased items on sites of schools, and had 9 biased items on types of schools and
7 biased items on sites of schools by Chi -Square analysis. The consistency of the bias
on both methods had 9 consistent items and 30 inconsistent items categorized into
types of schools, and had 7 consistent items and 33 inconsistent items categorized into
sites of schools. The second issue analyzed by Analysis of Variance had 9 biased items on
types of schools and 6 biased items on sites of schools, and had 8 biased items on
types of schools and 7 biased items on sites of schools by Chi -Square analysis. The
consistency of the bias on both methods had 7 consistent items and 30 inconsistent
items categorized into types of schools, and had 6 consistent items and 33 inconsistent
items categorized into sites of schools. English; there were 2 issues, each with 60 items.
The first issue analyzed by Analysis of Variance had 35 biased items on types of
schools and 4 biased items on sites of schools, and had 35 biased items on types of
schools and 6 biased items on sites of schools by Chi-Square analysis.
The consistency of the bias on both methods had 35 consistent items and 25
inconsistent items categorized into types of schools, and had 4 consistent items and 54
inconsistent items categorized into sites of schools. The second issue analyzed by
Analysis of Variance had 41 biased items on types of schools and 6 biased items on sites
of schools, and had 42 biased items on types of schools and 5 biased items on sites of
schools by Chi -Square analysis. The consistency of the bias on both methods had 41
consistent items and 18 inconsistent items categorized into types of schools, and had 5
consistent items and 54 inconsistent items categorized into sites of schools. Thai; there
were 2 issues, each with 80 items. The first issue analyzed by Analysis of Variance had
34 biased items on types of schools and 35 biased items on sites of schools, and had
36 biased items on both types of schools and sites of schools by Chi -Square analysis.
The consistency of the bias on both methods had 34 consistent items and 44
inconsistent items categorized into types of schools, and had 34 consistent items and
43 inconsistent items categorized into sites of schools. The second issue analyzed
by Analysis of Variance had 53 biased items on types of schools and 10 biased items
on sites of schools, and had 49 biased items on types of schools and 10 biased items on
sites of schools by Chi -Square analysis. The consistency of the bias on both methods had 46
consistent items and 24 inconsistent items categorized into types of schools, and had 8
consistent items and 68 inconsistent items categorized into sites of schools. Social
Studies; there were 2 issues, each with 100 items. The first issue analyzed by Analysis of
Variance had 39 biased items on types of schools and 17 biased items on sites of schools,
and had 39 biased items on types of schools and 15 bias items on sites of schools by
Chi -Square analysis. The consistency of the bias on both methods had 38 consistent
items and 60 inconsistent items categorized into types of schools, and had 14
consistent items and 82 inconsistent items categorized into sites of schools.
The second issue analyzed by Analysis of Variance had 56 biased items on types of schools and 33 biased items on sites of schools, and had 56 biased items on types of
schools and 36 biased items on sites of schools by Chi -Square analysis.
The consistency of the bias on both methods had 52 consistent items and 40
inconsistent items categorized into types of schools, and had 32 consistent items and
63 inconsistent items categorized into sites of schools. Physics; there were 2 issues, each
with 25 items. The first issue had 8 biased items on types of schools and 5 biased items
on sites of schools from both Analysis of Variance and Chi -Square analysis.
The consistency of the bias on both methods had 8 consistent items and 17
inconsistent items categorized into types of schools, and had 5 consistent items and 20
inconsistent items categorized into sites of schools. The second issue analyzed by
Analysis of Variance had 10 biased items on types of schools and 1 biased items on
sites of schools, and had 11 biased items on types of schools and 1 biased items on
sites of schools by Chi -Square analysis. The consistency of the bias on both methods
had 10 consistent items and 14 inconsistent items categorized into types of schools,
and had 1 consistent items and 24 inconsistent items categorized into sites of schools.
Chemistry; there were 2 issues, each with 40 items. The first issue analyzed by Analysis
of Variance had 10 biased items on types of schools and 8 biased items on sites of
schools, and had 11 biased items on types of schools and 8 bias items on sites of
schools by Chi –Square analysis. The consistency of the bias on both methods had 10
consistent items and 29 inconsistent items categorized into types of schools, and had
8 consistent items and 32 inconsistent items categorized into sites of schools.
The second issue had 8 biased items on types of schools and 3 biased items on sites of
schools from both Analysis of Variance and Chi –Square analysis. The consistency of the
bias on both methods had 8 consistent items and 32 inconsistent items categorized
into types of schools, and had 3 consistent items and 37 inconsistent items categorized
into sites of schools. Biology; there were 2 issues, each with 50 items. The first issue
analyzed by Analysis of Variance had 23 biased items on types of schools and 13 biased
items on sites of schools, and had 24 biased items on types of schools and 12 bias
items on sites of schools by Chi -Square analysis. The consistency of the bias on both
methods had 23 consistent items and 26 inconsistent items categorized into types of
schools, and had 12 consistent items and 37 inconsistent items categorized into sites of
schools. The second issue analyzed by Analysis of Variance had 19 biased items on types of schools and 11 biased items on sites of schools, and had 17 biased items on
types of schools and 11 biased items on sites of schools by Chi -Square analysis. The
consistency of the bias on both methods had 17 consistent items and 31 inconsistent
items categorized into types of schools, and had 11 consistent items and 39
inconsistent items categorized into sites of schools.
The result of the consistency analysis of the item bias analysis by using
Analysis of Variance and Chi-Square methods were found that the consistency of
detecting item bias on both methods of every test were significantly related at .05
level.