Makara Hubs-Asia, 2019, 23(1): 84-96

DOI: 10.7454/mssh.v23i1.1180718

Undergraduate Students’ Intelligence Profiles According to the Tes Intelligensi

Kolektip Indonesia Tinggi (TIKI-T): A Cluster Analysis Based on the Rasch
Model Person Ability

Whisnu Yudiana*, Heri Susanto, and Airin Triwahyuni

Faculty of Psychology, Universitas Padjadjaran, Jatinangor 45363, Indonesia

*E-mail: whisnu.yudiana@unpad.ac.id

Abstract
This research aims to identify undergraduate students’ intelligence profiles using a two-stage cluster analysis based on the
person’s ability of the Rasch model to examine the effect of the clusters on academic performance. A total of 1443
undergraduate students from nine academic disciplines at Universitas Padjadjaran in Bandung, Indonesia, participated in
the study, completing 11 subtests of the Tes Intelligensi Kolektip Indonesia Tinggi (TIKI-T). A hierarchical cluster
analysis approach using Ward’s linkage method and squared Euclidean distance was conducted, followed by a
nonhierarchical k-means cluster analysis using simple Euclidean distance as the similarity measure to examine two-, three-,
four-, and five-cluster solutions. An intra-class correlation (ICC) and a discriminant analysis were also conducted to
validate the cluster membership results. This research identified five profiles of intelligence that had an effect on academic
performance. Students with high scores in the scholastic aptitude subtests tended to have higher grade point average than
those with high scores in the nonverbal ability subtests and the speed and accuracy ability subtests. The findings can be
used as a recommendation for psychologists in Indonesia for university placement tests.

Profil Kecerdasan Mahasiswa Berdasarkan Tes Intelligensi Kolektip Indonesia Tinggi

(TIKI-T): Analisis Klaster berdasarkan Person Ability dari Model Rasch

Abstrak

Penelitian ini bertujuan untuk mengidentifikasi profil kecerdasan dari mahasiswa dengan menggunakan dua tahap
analisis klaster berdasarkan person ability dari model pengukuran Rasch dan melihat dampak dari profil tersebut
terhadap prestasi akademik. Sebanyak 1443 mahasiswa dari sembilan program studi di Universitas Padjadjaran di
Bandung Indonesia menyelesaikan sebelas kelompok soal dari TIKI-T. Pendekatan Hierarki dengan metode Ward’s
Linkage and Square Euclidean distance dilakukan, diikuti oleh non-hierarki analisis klaster dengan dua, tiga, empat dan
lima solusi klaster sebagai pembanding. Analisis intra-class correlation (ICC) dan analisis diskriminan juga digunakan
untuk memvalidasi keanggotaan dari setiap klaster. Penelitian ini mengidentifikasi lima profil kecerdasan yang
berpengaruh terhadap prestasi akademik. Ditemukan bahwa mahasiswa dengan nilai yang tinggi dalam kelompok soal
yang mengukur kemampuan skolastik cenderung mendapatkan hasil yang lebih baik dalam prestasi akademik
dibandingkan dengan mahasiswa yang memiliki nilai tinggi dalam kelompok soal yang mengukur kemampuan
nonverbal ataupun tinggi dalam kelompok soal yang mengukur kecepatan dan ketepatan. Temuan ini dapat digunakan
sebagai rekomendasi bagi psikolog di Indonesia untuk tes penempatan universitas.

Keywords: academic performance, cluster analysis, intelligence profile, the TIKI-T

Citation:
Yudiana, W., Susanto, H., & Triwahyuni, A. (2019). The intelligence profile of the university students using the Tes
Intelligensi Kolektip Indonesia Tinggi (TIKI-T): A cluster analysis based on the rasch model person ability. Makara Human
Behavior Studies in Asia, 23(1), 84-96. doi: 10.7454/hubs.asia.1180718

1. Introduction educational and professional success (Buschkuehl &

Jaeggi, 2010; Rosander, Bäckström, & Stenberg, 2011;
Intelligence is related to performance in a wide range of Schmidt & Hunter, 1998). Research in educational
cognitive tasks and is one of the best predictors for settings has revealed that intellectual factors obtained

84 July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 85

from psychometric tests have medium to high intelligence profiles based on nine subtests in the
correlations with students’ academic performance. For Intelligenz Struktur Test from people who were
instance, Mackintosh (1998) found that the correlation successful in various kinds of jobs; Letteri (1980)
between IQ and school performance grade ranged proposed a cognitive profile that related to seven
from 0.40 to 0.70. Additionally, a national survey dimensions of the cognitive test for seventh and eighth
conducted by Deary, Thorpe, Wilson, Starr, and grade students; Kim, Frisby, and Davison (2004)
Whalley (2003) in the United Kingdom discovered that explored the cognitive profiles of Woodcock–Johnson
general intelligence had an enormous contribution to 25 Psychoedu-cational Battery; and both Bergeron and
academic subjects for 11-year-old students with Floyd (2013) and Mayes and Calhoun (2004)
correlations ranging from 0.43 to 0.77. However, the investigated the cognitive profiles of children with
relationship was higher when the latent intelligence disabilities.
traits (Spearman’s g of a psychometric test) and the
latent traits of educational performance were correlated. In the Indonesian population, Djunaidi and Suryabrata
(1997) investigated intelligence profiles based on 11
The range of findings in intelligence research including subtests of the Tes Intelligensi Kolektip Indonesia
university students have tended to relate to the level of Tinggi (TIKI-T) (Drenth, Dengah, Bleichrodt, Soemarto,
the correlation coefficient. Busato, Prins, Elshout, and & Poespadibrata, 1977) from a university student
Hamaker (2000), for instance, found that the minimum sample in Yogyakarta, Indonesia. The 11 subtests
correlation between intelligence and academic measured several constructs of cognitive abilities, and
performance in psychology students was only 0.13. A the results identified four intelligence profiles of the
light to moderate relationship was also found in general university students. The first profile had a high
university students (Komarraju, Ramsey, & Rinella, contribution from the visualization, components, hidden
2013). It should be emphasized that the correlation figures, number series, and spatial orientation subtests.
between IQ and academic performance for university The accuracy and speed, arithmetic, and verbal analogies
students was lower than that for junior and elementary subtests highly contributed to the second profile. The
school students. This trend was related to the restriction third profile was influenced by the figure classification
range of university students who are selected on more subtest. Finally, the fourth profile was only influenced
specific criteria (Chamorro-Premuzic & Furnham, 2005). by the word relations subtest. Nevertheless, no evidence
was found that related the effect of the intelligence profile
The findings mentioned above focus on the global to academic performance, and the analysis was based on
capacity of the intellect in academic performance, the calculated scores that used the classical test theory
otherwise known as “g factor” (Spearman, 1904). approach.
However, a lot of research has been conducted to
discover the specific factors, or “s-factors,” that Similar to the above mentioned study (Djunaidi &
influence students’ attainment in school such as Suryabrata, 1997), this research is also focused on
working memory (Weber, Lu, Shi, & Spinath, 2013), university students’ intelligence profiles according to
numerical ability (Halberda, Mazzocco, & Feigenson, the TIKI-T (Drenth et al., 1977). The most appropriate
2008), and verbal ability (Lee, Ng, Ng, & Lim, 2004). procedure to define natural groupings within data is
More specifically, research has tended to focus on one through a cluster analysis (Hair, Black, Babin, Anderson,
subject only (e.g., mathematics) instead of academic & Tatham, 2014; Mooi & Sarstedt, 2011) which is a
performance as a whole (Landerl, Bevan, & Butterworth, multivariate procedure to identify groupings within
2004; Landerl & Moll, 2010). data. A cluster analysis of variables resembles a factor
analysis because both procedures detect related groups
Another approach to predict academic performance of variables (Hair et al., 2014). In this study, cluster
from intelligence is the use of cognitive profiles. A analysis was used to identify the naturally occurring
profile approach provides an alternative way of profiles or groups of students within a sample that has a
representing students’ intelligence by including similar pattern of scores on the 11 subtests of the TIKI-
information about the combinations of multiple T. Unlike previous study, however, the present study
constructs and the constructs’ respective magnitudes. applies a Rasch Model analysis to the calculation of the
By using a profile approach, it is possible to identify scores (Rasch, 1960).
which profiles relate to high academic performance and
which relate to low academic performance (Letteri, The Rasch analysis, originally developed by George
1980). A profile approach can thus help identify Rasch, is a part of the item response theory. Rasch
possible learning deficits. (1960) proposed a mathematical model based on the
relationship between the probabilities of a student’s
To identify profiles based on intelligence factors, several response to an item as a function of the student’s
studies have been conducted in both educational and “ability.” The Rasch model for dichotomous data is
occupational contexts: Amthauer (1970) identified appropriate to analyze multiple choice items that are

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

86 Yudiana, Susanto, & Triwahyuni

scored as either right or wrong. The Rasch analysis from the students as an agreement for their
provides more advantages than the classical test theory participation in the study. Third, the data was collected
(CTT) employed in Djunaidi and Suryabrata’s (1997) in a classroom setting with 60–100 students in a room.
study. One limitation of CTT is that the item and person The tests were administered by experienced instructors,
characteristics are dependent on each other. This means with the assistance of three to five co-instructors. All
that item parameters (e.g., item difficulty) might change the processes were conducted under the supervision of
depending on the subpopulation considered in the study psychologists. The time used for a single administration
(i.e., test items could appear easy when the test is was 90–120 min. Lastly, the participants received a
administered on a high ability subpopulation, and vice reward after finishing the test.
versa). The Rasch model analysis overcomes this
limitation by allowing the formulation of item Measures. The primary data used in this study was
characteristics and the personal abilities by distinctive obtained from the TIKI-T. The TIKI-T consists of 11
parameters, such as sample invariants (Bond & Fox, subtests in a multiple-choice format with four to five
2015). The Rasch analysis also provides a transformation options. Each subtest measures a different construct,
of an ordinal score into an interval level variable where and each has a different number of items and time
valid scores and access to parametric statistics are limits. Based on the test manual (Drenth et al., 1977),
required (Tennant & Conaghan, 2007). Another the test specifications for each subtest are as follows:
advantage of the Rasch analysis is related to its • Arithmetic (a = 0.96). This subtest consists of 40
assumption of unidimensionality: by assuring the items that measure the ability to solve simple
unidimensionality of the test, its construct validity will numerical problems that require arithmetic operations,
be supported, allowing for adequate theoretical such as addition, subtraction, multiplication, and
interpretations of the test score (Van der Ven & Ellis, division. The time limit for this subtest is 7 min.
2000). The Rasch model has been widely used to • Component (a = 0.88). This subtest consists of 26
analyze and develop several cognitive tests (Ariffin et items that measure the ability to manipulate and
al., 2010; Freitas, Prieto, Simões, & Santana, 2014; transform figural material. The time limit for this
Koski, Xie, & Finch, 2009; Primi, 2014; Woodcock, subtest is 7 min.
1999; Woodcock, McGrew, & Mather, 2001). The • Word relations (a = 0.90). This subtest consists of
research questions for the current study were therefore: 40 items that measure the ability to identify two
(1) What are university students’ intelligence profiles words with either identical or contrasting meanings.
based on the Rasch person ability? (2) Does academic The time limit for this subtest is 5 min.
performance differ based on the intelligence profile?
• Figure classification (a = 0.81). This subtest
consists of 30 items that measure the ability to
2. Methods classify figural objects. The time limit for this
subtest is 12 min.
Sample and Participants. 1443 undergraduate students
from 9 academic disciplines at Universitas Padjadjaran • Number series (a = 0.64). The subtest consists of 20
in Bandung, Indonesia, participated in the study. The items that measure the ability to find a number in a
study programs were randomly selected from more than series of numbers arranged according to a certain
50 disciplines. There were 338 animal husbandry principle. The time limit for this subtest is 10 min.
students (23.4%), 266 psychology students (18.4%), • Accuracy and speed (a = 0.88). The subtest consists
206 Sundanese literature students (14.3%), 181 biology of 100 items. The test requires the speed and
students (12.5%), 158 pharmacy students (10.9%), 91 accuracy to identify identical or different words.
geological engineering students (6.3%), and 76 The time limit for this subtest is 4 min.
international relations students (5.3%); the remaining • Visualization (a = 0.84). This subtest composes 30
students were studying library sciences (4.5%) and items that measure the ability to visualize three-
sociology (4.5%). The majority of the participants were dimensional figures from two-dimensional figures.
first-year (40%) and second-year (48%) students, and The time limit for this subtest is 5 min.
only 116 (1.50%) were third-year students. According • Spatial orientation (a = 0.98). The subtest consists
to gender classification, more than half of the of 40 items that measure the ability to rotate an
participants were women, 966 (66.9%), and the rest object and reflection from the initial figure. The
were males, 477 (33.10 %). The mean age of the time limit for this subtest is 10 min.
participants was 19.34 years (SD=0.91). • Verbal analogies (a = 0.53). The subtest consists of
18 items that measure deductive reasoning. The
Participation in the study was voluntary. Several time limit for this subtest is 4 min.
procedures were conducted before collecting the data. • Hidden figure (a = 0.76). The test contains 20 items
First, the researcher explained the aim of the study and that measure the ability to search a distracting
the process of the data collection to the target students. perceptual field to find a given configuration. The
Second, the informed consent forms were collected time limit for this subtest is 6 min.

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 87

• Word composition (a = not reported). The test coefficient greater than 0.5. The analyses for the Rasch
contains 60 items that measure the ability to find the model in this study used Conquest software. Except for
missing letters of a frequently used word. The time examining PCA and local items dependence, the
limit for this subtest is 4 min. Winstep software was used, as this analysis was
available in Conquest software.
The secondary data collected was related to academic
performance (grade point average—GPA). The range of Secondly, screenings for univariate and multivariate
the students’ GPA was from 0.00 to 4.00. An analysis of normality and outliers were conducted. Descriptive
performance (ANOVA) found that there was a statistics (i.e., mean, standard deviations, and bivariate
significant difference in average GPA among the nine correlations) were analyzed to provide a description of
academic disciplines. Therefore, for the analysis, the the sample. Bivariate correlations were computed because
GPA scores were converted into Z-scores based on the they were important to examine multicollinearities. Hair
mean and standard deviations for each academic et al. (2014) explained the effect of multicollinearity to
discipline. the cluster analysis as a form of implicit weight. To
examine multicollinearity, a review of the correlation
Analyses. Several procedures were used to analyze the matrix was conducted. According to Field (2009),
data. Firstly, the Rasch model analysis for dichotomous correlation coefficients between variables should not
data was conducted to acquire the person ability scores. exceed 0.80. The results are shown in Table 1.
This model was appropriate for analyzing the multiple-
choice items which are scored as either right or wrong Thirdly, the exploration of the cluster from the data was
(Rasch, 1960). The person ability scores represent an conducted using a two-stage cluster analysis. The first
estimate of a person’s underlying ability related to stage is the hierarchical cluster analysis using Ward’s
his/her performance on a set of items that measure latent method and squared Euclidean distance. Ward’s
traits (Bond & Fox, 2015). In the analysis, person ability minimum variance method produced the most reliable
estimates were computed using weighted likelihood results. Ward’s algorithm has been commonly used in
estimates (WLEs) for dichotomous responses (Warm, intelligence studies because it has performed quite well
1989). The Rasch model analysis has several important in comparison to alternative clustering methods
requirements, specifically model fit or item fit, (Donders, 1996; Hale, Casey, & Ricciardi, 2014). This
unidimensionality, and local dependency (Bond & Fox, method is appropriate to obtain the same approximate
2015). size of samples as the cluster group. The range of
profiles was explored by reviewing the agglomeration
Two scores were used to evaluate the item fit: Infit schedule coefficient from two-, three-, four-, or five-
mean square (MNSQ) and outfit MNSQ. However, cluster solutions which is acceptable to interpret. The
outfit MNSQ is an unweighted score which is more process continued with a nonhierarchical k-means
sensitive to outlier responses (Bond & Fox, 2015), and cluster analysis using simple Euclidean distance. This k-
therefore in the present analysis only infit MNSQ values means specified a two-, three-, four-, or five-cluster
were used to evaluate the misfit items. Items with infit solution. This approach is recommended to correct
MNSQ statistics between 0.75 and 1.30 were included fusion errors that random cluster centers can produce
in the next analysis (Bond & Fox, 2015). The (Hair et al., 2014). To review the reliability, the cluster
dimensionality of the TIKI-T was assessed using a membership by hierarchical and k-means, an intraclass
principal component analysis (PCA) of the residuals. correlation coefficient (ICC), was computed to acquire
According to Linacre (2006), variance explained by the index of cluster “stability” (reliability) of the two
measures of more than 60% and unexplained variance methods (Morris, Blashfield, & Satz, 1981). Thus, all
by first contrast of less than 5% could be taken as initial prospective solutions were compared on the basis of
indicators of unidimensionality. However, Sumintono association between the results from the hierarchical
and Widhiarso (2015) recommended another criterion of and the k-means.
dimensionality: a subtest with more than 40% variance
explained by measures and less than 15% unexplained Cohen’s Kappa statistics and one-way random effects
variance is accepted as having unidimensionality, ICC were used to compare the solutions in terms of
whereas more than 60% variance explained by measures membership agreement and profile similarity,
is considered as having excellent unidimensionality. respectively. Higher reliability index of the cluster
Lastly, local dependence is important to evaluate the solutions was chosen for further analysis. For the
independency of one datum in relation to another external validation, this study used discriminant
datum, that is, the value of one datum has no influence analysis, also known as supervised classification of
on another datum. High correlation, whether positive or some observations, to classify others (Hair et al., 2014).
negative, is an indicator of local dependency for any This method was used to assess whether or not a set of
pair of items (Wright, 1996). Cohen (1988) stated that variables discriminates between the cluster groups by k-
high correlation can be indicated by a correlation means method. Discriminant function coefficients as the

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

88 Yudiana, Susanto, & Triwahyuni

Table 1. Descriptive Statistics and Correlation Matrices for the Study Variables (N = 1368)

Subtest 1 2 3 4 5 6 7 8 9 10 11
1 AR 1
2 CO 0.263 1
3 WR 0.317 0.285 1
4 FC 0.443 0.438 0.379 1
5 NS 0.514 0.305 0.329 0.466 1
6 AS 0.206 0.194 0.146 0.203 0.172 1
7 VI 0.337 0.464 0.293 0.510 0.358 0.193 1
8 SO 0.368 0.355 0.282 0.494 0.406 0.143 0.431 1
9 VA 0.291 0.205 0.260 0.380 0.343 0.147 0.286 0.301 1
10 HF 0.314 0.364 0.254 0.458 0.333 0.161 0.388 0.367 0.265 1
11 WC 0.259 0.237 0.229 0.280 0.261 0.168 0.226 0.254 0.189 0.251 1

Original
40 26 40 30 20 100 30 40 18 20 60
N-item
Mean
(WLE 1.674 1.527 1.891 0.974 0.320 3.894 1.680 0.748 0.153 1.194 3.690
score)
SD 0.937 1.104 0.934 1.385 1.359 1.176 1.111 2.055 0.857 1.488 1.039
Note: AR, arithmetic; CO, component; WR, word relation; FC, figure classification; NS, number series; AS, accuracy and speed;
VI, visualization; SO, spatial orientation; VA, verbal analogy; HF, hidden figure; WC, word composition. Values in italics indicate p
< 0.01 (two-tailed). *p<0,05; **p < 0,01

results from the discriminant analysis showed the best composition (23%, 14 items) had items with infit
predictor on the dependent variable. The dependent MNSQ outside the accepted range. All those items were
variable for this analysis was the clustered group, while excluded from the following analysis. The data show
the independent variables were 11 subtests of the TIKI- that “accuracy and speed” and “word composition” had
T. The coefficients were standardized to remove the the highest number of items excluded from the subtest.
effects of differences between mean and standard These subtests measure speed ability: how fast a person
deviations. does a simple task using cognitive automatic processing.
The analysis for such a subtest should be based on the
The final analysis for this study focused on which frequency distribution rather than solely using the item
profiles were the best predictor of academic performance. analysis.
A one-way ANOVA was conducted using the k-means
cluster analysis profile groups as the independent Table 3 shows that all subtests had more than 40%
variable and the standardized GPA as the dependent variance explained by measures, which ranged between
variable. 40% (verbal analogies) and 82% (accuracy and speed).
All subtests had less than 15% unexplained variance on
3. Results first contrast, which ranged between 2% (word
composition) and 6% (number series). This means that all
Rasch analysis. The Rasch model analysis for the item fit TIKI-T subtests meet the unidimensionality requirement.
is presented in Table 2. The better an item was, the closer Meanwhile, according to Wright (1996) and Cohen
it was to the model fit. The results show that more than (1988)’s criteria for local item dependence, only two
half of the TIKI-T subtests had items with infit statistics subtests “accuracy and speed” and “word composition”
within the accepted range 0.75 and 1.30 (Bond and Fox, had items with correlation coefficient greater than 0.50.
2015). Only arithmetic (8%, three items), figure classifi-
cation (7%, two items), accuracy and speed (66%, 66 Both subtests had a high percentage variance explained
items), spatial orientation (5%, two items), and word (82% and 72%, respectively), yet showing high local
Table 2. Infit and Outfit Statistics from Rasch Model Analysis

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 89

Infit MNSQ Outfit MNSQ

% of items
Subtest
excluded
Mean SD Min Max Mean SD Min Max

Arithmetic 8 0.97 0.17 0.71 1.40 2.29 1.79 0.54 6.74

Component 0 0.98 0.07 0.84 1.14 1.32 0.47 0.83 2.77

Word relation 0 0.99 0.11 0.76 1.29 1.10 0.39 0.64 2.47

Figure classification 7 0.99 0.14 0.72 1.43 1.08 0.38 0.60 2.21

Number series 0 1.00 0.09 0.85 1.13 1.35 0.81 0.79 3.69

Accuracy and speed 66 0.89 0.33 0.53 2.63 6.35 4.10 0.36 9.90

Visualization 0 0.98 0.11 0.78 1.17 1.28 0.67 0.45 3.76

Spatial orientation 5 0.96 0.14 0.75 1.32 1.98 1.81 0.68 9.90

Verbal analogies 0 1.00 0.06 0.91 1.13 1.02 0.19 0.72 1.52

Hidden figures 0 1.00 0.10 1.00 1.27 1.14 0.34 1.14 1.90

Word composition 23 0.94 0.20 0.61 1.49 4.42 3.67 0.36 9.90

Table 3. Dimensionality and Local Dependence Analysis

Dimensionality Local dependence

Subtest % unexplained variance Largest standardized
% variance Indication of local
in first contrast residual correlation
explained dependence
range
Arithmetic 63 3 0.37 to 0.23 No

Component 51 5 -0.19 to 0.40 No

Word relation 53 4 0.31 to 0.22 No

Figure classification 41 3 -0.15 to 0.21 No
Number series 40 6 -0.21 to 0.16 No
Accuracy and speed 82 3 1.00 to 0.93 Yes
Visualization 53 4 -0.17 to 0.15 No
Spatial orientation 59 3 0.36 to 0.20 No

Verbal analogies 40 5 -0.19 to -0.13 No

Hidden figures 45 5 -0.12 to 0.12 No

Word composition 72 2 0.81 to 0.39 Yes

dependency. The effect of item independence is inflating Descriptive statistics. Data screening procedures did not
the ability estimates of items at a given scale. Moreover, identify any variables as non-normal (skewness/kurtosis
it seriously distorts the qualities of the items (Sideridis, > 2), but 65 univariate outliers were found (z > ±3.0).
2011). However, in this research, items with high Therefore, those cases were excluded for the following
correlations were also misfit items. Thus, such items were analysis. This is important because outliers can have a
excluded in cluster analysis. significant impact on the results, particularly in cluster
analysis (Hair et al., 2014).
Table 4. Group Descriptive Statistic and Test Equality for Estimation Sample in the Three-Group Discriminant Analysis

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

90 Yudiana, Susanto, & Triwahyuni

Dependent variable group mean

Test of equality of group Means
(Z-score from the WLE)
Independent variable 1st
2nd
Cluster 3rd Cluster 4th Cluster 5th Cluster Wilks’
Cluster (n F Sig.
(n = (n = 300) (n = 135) (n = 464) Lambda*
= 283)
186)
Arithmetic -0.97 0.70 0.08 -0.92 0.34 0.60 226.22 0.00
Component 0.08 0.86 -0.57 -1.27 0.28 0.56 268.22 0.00
Word relation -0.48 0.68 0.00 -0.96 0.23 0.73 125.59 0.00
Figure classification -0.79 0.94 -0.25 -1.25 0.41 0.46 404.65 0.00
Number series -0.95 0.75 -0.03 -1.03 0.37 0.58 251.41 0.00
Accuracy and speed 0.07 0.60 -0.16 -0.69 0.01 0.86 55.01 0.00
Visualization -0.31 0.90 -0.33 -1.37 0.29 0.55 281.54 0.00
Spatial orientation -0.59 0.83 -0.39 -1.21 0.42 0.57 261.75 0.00
Verbal analogies -0.68 0.56 0.13 -0.89 0.24 0.73 128.07 0.00
Hidden figures -0.39 1.05 -0.34 -1.00 0.12 0.61 217.44 0.00
Word composition -0.04 0.63 -0.43 -0.88 0.28 0.76 109.52 0.00

Note: *Wilks’ Lambda (U-statistic) and univariate F ratio with 4 and 1363 degrees of freedom

1.5

0.5

0
AR CO WR FC NS AS VI SO VA HF WC

-0.5

-1

-1.5

1St Cluster 2nd Cluster 3rd Cluster 4th Cluster 5

Figure 1. Results of the k-means Cluster Analysis using WLE Scores. AR, Arithmetic; CO, Component; WR, Word Relation;
FC, Figure Classification; NS, Number Series; AS, Accuracy and Speed; VA, Visualization; SO, Spatial
Orientation; VA, Verbal Analogy; HF, Hidden Figure; WC, Word Composition

by comparing the profiles acquired from the initial

Cluster analysis. Cluster analysis. The reliability of the Ward’s analysis (first stage) to those derived from the k-
three-, four-, and five-cluster solutions were evaluated means analysis (second step). The ICC for the five-

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 91

cluster solution was -0.679 (p = 1.00), for the four- subtests. The fourth profile (bold dashed line) was
cluster solution was 0.164 (p < 0.05), and for the three- labeled as low level of intelligence factors (n = 135) as it
cluster solution was 0.405 (p < 0.05). These results was visualized by the lowest intelligence score for every
indicated that the three- and four-cluster solutions were subtest. The fifth profile (dashed-dotted line) was
less stable, whereas the five-factor solution was the labeled as above average level of intelligence (n = 464)
most stable. Therefore, the five-cluster solution was as this profile was characterized by a relatively high
chosen for a discriminant analysis. level of all intelligence factors. Almost one-third of the
samples were classified in this profile.
The group descriptive statistic and test equality are
presented in Table 4. Significant mean differences were Table 5 presents the membership proportion for the nine
observed for all the predictors on the dependent academic disciplines. Overall, the disciplines were
variable. The Z-score from WLE person ability for the classified into three major groups based on the
second cluster was the highest for all independent proportion in each cluster. The first group were
variables. It indicated that the students in the second psychology, pharmacy, and geological engineering
cluster had the highest intelligence factor. The log students. In these academic disciplines, more than 65%
determinants were quite similar between -5.532 and of the students were categorized as followed by average
6.567. The Box’s M indicated that the assumption of level of intelligence, and only a few portions of the
equality of covariance matrices was violated. However, students with a low level of intelligence factors. The
because the sample size was large, this was not an second group were biology, animal husbandry,
indication of a problem. The discriminate function international relations, and sociology students. The
revealed a significant association between groups and majority of the students in these academic disciplines
all predictors. A canonical correlation of 0.938 suggests were classified as possessing an average and above
that the model explains 87.9% of the variation in the average level of intelligence factors (55–60%), followed
grouping variable, that is, whether the student grouped by high level of intelligence (10–25%), and the rest as a
into the first, second, third, fourth, or fifth cluster. The low and below average level of intelligence factor. The
cross-validate classification showed that 94.4% were third group were composed of students studying library
correctly classified, which indicated that the classification science and Sundanese literature. In these disciplines,
from the cluster analysis process was valid. 55–60% students were categorized as possessing low
and below average level of intelligence factors,
Description of clusters. The five clusters generated followed by average ability students, and less than 10%
based on the k-means analysis were assigned descriptive of students were categorized as possessing a high level
labels reflecting the most salient features of each mean of intelligence factors.
(WLE) from the TIKI-T profile. Figure 1 illustrates the
mean of each subtest on the TIKI-T for every cluster Intelligence profile and academic performance. A
profile using the standardized WLE scores. This one-way between-groups analysis of variance was
information was provided to aid the interpretation and conducted to explore the impact of intelligence profiles
labeling of the profiles. The labels were used to on GPA and IQ deviation. As mentioned previously, the
characterize the profiles relative to their counterparts. In GPA was standardized to minimize the effect of mean
this case, they mostly corresponded to high and low differences between the academic disciplines.
levels of the students on each subtest. From Figure 1,
the first profile (dotted line) was labeled as below Meanwhile, IQ deviation was a standardized score with
average level of intelligence factors (n = 186) as it was the mean 100 and SD 15. Participants were divided into
generally represented by a level of intelligence profile five groups according to the level of intelligence profile:
between 0 and -1. This profile had the highest score in Group 1, low level of intelligence; Group 2, below
the component subtest, speed and accuracy subtest, and average level of intelligence; Group 3, average level of
word composition subtest. The second profile (straight intelligence; Group 4, above average level of intelligence;
line) was labeled as high level of intelligence factors (n and Group 5, high level of intelligence there was a
= 283) as this profile was characterized by a high level statistically significant difference at the p < 0.05 level in
of all intelligence factors. The third (dashed line) was standardized GPA scores for the three groups
labeled as average level of intelligence factors. The intelligence profiles: F (4, 1342) = 6.425, p < 0.01.
third profile (dashed line) was labeled as average level Despite reaching statistical significance, the actual
of intelligence factors (n = 300) as it was generally difference in the mean scores between the groups was
represented by intelligence ability that was relatively quite small. The effect size, calculated using etasquared,
neither too high nor too low compared to the other was only 0.019. With equal variance not assumed, post-
clusters. This profile was also characterized by the hoc comparisons using Dunnett’ C-test indicated that the
highest score on both the arithmetic and visualization
Table 5. The Membership Proportion for Each Cluster Based on Academic Discipline

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

92 Yudiana, Susanto, & Triwahyuni

Study program
Cluster 1 2 3 4 5 6 7 8 9 Total
1 (%) 3.10 3.20 6.30 10.30 13.90 16.40 24.10 27.10 37.20 13.60

2 (%) 35.20 30.40 23.30 27.60 12.70 26.00 10.30 8.50 4.40 20.70

3 (%) 21.10 21.50 29.50 17.20 26.30 20.50 19.00 11.90 15.60 21.90

4 (%) 1.90 3.20 7.40 3.40 11.40 0.00 19.00 27.10 25.60 9.90

5 (%) 38.70 41.80 33.50 41.40 35.80 37.00 27.60 25.40 17.20 33.90

Note. Academic discipline: 1, psychology; 2, pharmacy; 3, biology; 4, geological engineering; 5, animal husbandry; 6, international
relations; 7, sociology; 8, library science; 9, Sundanese literature.

Figure 2. The Relation of Intelligence Profile with GPA and IQ

mean score for Group 5 (M = 0.170, SD = 0.884) was 1.060), Group 2 (M = -0.161, SD = 1.168), and Group 3
significantly different from Group 1 (M = -0.248, SD = (M = -0.049, SD = 0.971) and did not differ significantly

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 93

with Group 4 (M = 0.090, SD = 0.96). There was a tended to have higher GPAs than those with high
significant difference between Group 4 with Groups 1 nonverbal abilities and high speed and accuracy abilities.
and 2. There was no significant difference found between
Group 2 and Group 3; between Group 2 and Group 4; and This study has several implications for psychologists
between Group 3 and Group 4. Details of mean who use the TIKI-T as an instrument for student
differences are shown in Figure 2. placement or career guidance. First, the study has
demonstrated the use of cluster analysis for investigating
4. Discussion the intelligence profiles of the TIKI-T. From the cluster
analysis, it should be highlighted that the cluster of the
The results suggest that reliable intelligence profile TIKI-T did not divide students based on their academic
patterns of the TIKI-T subtest scores can be derived discipline due to the similar proportion of students in
using cluster analysis for undergraduate students. This each cluster. Moreover, no specific intelligence profile
study found that students from the nine academic was found for each academic discipline. However,
disciplines showed variations in both level and pattern psychologists may find that the intelligence profiles
of performance across the TIKI-T subtests. Although found in this study can be used as a basis for the
previous research discovered four clusters in the prediction of students’ academic performance. To
undergraduate student samples (Djunaidi & Suryabrata, identify students’ abilities, psychologists should take a
1997), there were five clusters obtained in the present look at intelligence profiles based on the score of
investigation based on k-means clusters analysis, with individuals and categorize them into high level of
the highest ICC compared to three- and four-cluster intelligence, average level of intelligence, and low level
solutions. The five clusters were highly correlated with of intelligence. Students with a high level of intelligence
IQ deviations. This finding explained a high correlation have a good opportunity to succeed in academic
between g factor and s factors (Spearman, 1927). In disciplines with high learning demands. In the present
other words, the level of the subtest scores was a research, these disciplines included psychology, geology
reflection of the IQ deviation. engineering, and pharmacy. On the other hand,
individuals with a low level of intelligence would be
The terms used in the labels of the clusters were similar better suited to a less demanding discipline for a greater
to the levels of intelligence in the Differential Ability chance of good performance. Moreover, the identification
Scales Second Edition (DAS-II; Urbina, 2011). In DAS of students with lower TIKI-T scores may be useful for
II, the categorizations are low (IQ level 70–79), below designing learning interventions and preventing further
average (80–89), average (90–109), above average academic problems. Besides, psychologists also have to
(109–120), and high (120–129). In this research, there consider the scholastic aptitude scores of each individual
was a slight difference in the level of IQ for the in making a decision as this score contributes to the
categories above average and high. The majority of the GPA score.
students could be grouped into above average level of
intelligence (33.90%), with the top three subtests being This study has some limitations that need to be
figure classification, number series, and spatial orienta- highlighted. First, the use of the cluster analytic method.
tion which measure nonverbal reasoning abilities Even though multiple methods were used to validate the
(Drenth et al., 1977). The second largest majority of cluster membership, cluster analysis represents a
students were clustered into the average level of relatively subjective research tool (Lange, Iverson,
intelligence (21.90%), with the top three subtests being Senior, & Chelune, 2002). The results from the analysis
arithmetic, verbal analogies, and word relations which will be different each time a different sample is used.
mainly measure the general scholastic aptitude (Drenth Second, regarding the similarity coefficient, the method
et al., 1977). Although a large majority of students were of distance was used in this research, and association
grouped into high level of intelligence (33.90%), the indices followed conventional standards and were
hidden figure, figure classification, and visualization empirically driven. In the end, the researchers’
subtests were shown as the top three subtests. In the knowledge and theoretical background should be
below average level of intelligence group (13.60%), the adequate (Hale et al., 2014). Third, this study used
top two subtests were word composition and accuracy person ability estimates from the Rasch model.
subtests. These subtests measured speed and accuracy However, in the analysis, the scores were transformed
abilities (Drenth et al., 1977). The cluster with the into Z-scores to justify that all the variables had the
smallest proportion of students was the low level of same distance. Moreover, several data were also
intelligence (9.90%). Similar to the below average level indicated as outliers. Therefore, the results from the
of intelligence group, the top two subtests were word Rasch analysis was approximately normal and unbiased
composition and accuracy subtests which measure speed except for the extreme abilities, which are consistently
and accuracy abilities (Drenth et al., 1977). Based on biased toward the mean by a floor and ceiling effect
these results, the students with high scholastic aptitude (Pelton, 2002). A comparative study needs to be
conducted to justify the different results between the

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

94 Yudiana, Susanto, & Triwahyuni

Rasch model and CTT. Finally, this study provides a

basis for future research on the interpretation of the Chamorro-Premuzic, T., & Furnham, A. (2005).
TIKI-T intelligence profiles. The findings of this study Personality and intellectual competence. Mahwah, NJ:
do not provide evidence for the use of TIKI-T Lawrence Erlbaum Associates, Publishers.
intelligence profile for the prediction of student
performance across different academic disciplines, thus Cohen, J. (1988). Set correlation and contingency
further study may be needed to provide the basis for tables. Applied Psychological Measurement, 12(4),
profile interpretation of the TIKI-T according to specific 425–434. 10.1177/014662168801200410.
academic disciplines.
Deary, I. J., Thorpe, G., Wilson, V., Starr, J. M., &
5. Conclusion Whalley, L. J. (2003). Population sex differences in IQ at
age 11: The Scottish mental survey 1932. Intelligence,
This paper contributed to the literature by outlining the 31(6), 533–542. doi: 10.1016/S0160-2896(03)00053-9.
two stages of cluster analysis: the first step was
hierarchical and the second step was a k-mean cluster Djunaidi, A., & Suryabrata, S. (1997). Studi mengenai
analysis for profiling intelligence and analyzing the profil inteligensi mahasiswa UGM berdasarkan TIKI-T
effect of the profile to academic performance. The two (A study of UGM students’ intelligence profile based on
methods used were ICC and discriminant analysis TIKI-T). Yogyakarta: Universitas Gadjah Mada.
which were conducted to justify the membership of the
respondents. This paper also contributed to the Donders, J. (1996). Cluster subtypes in the WISC-III
identification of which factor in the intelligence profile standardization sample: Analysis of factor index scores.
contributed to academic performance. These results may Psychological Assessment, 8(3), 312–318. doi: 10.1037/
be used by psychologists to identify and predict 1040-3590.8.3.312.
undergraduate candidates’ academic performance.
Drenth, P. J. D., Dengah, B., Bleichrodt, N., Soemarto,
Declaration of Interest & Poespadibrata, S. (1977). Test intelligensi kolektip
Indonesia. Lisse: Swets & Zeitlinger.
The authors report no conflicts of interest in this work.
Field, A. (2009). Discovering statistics using SPSS.
References London: Sage publications.

Amthauer, R. (1970). Intelligence structure test. Freitas, S., Prieto, G., Simões, M. R., & Santana, I.
Göttingen, Germany: Hogrefe. (2014). Psychometric properties of the Montreal
Cognitive Assessment (MoCA): an analysis using the
Ariffin, S. R., Asari, S. M., Mohamed, S., Shahar, S. N., Rasch model. The Clinical Neuropsychologist, 28(1),
Ishak, N. M., Din, R., … Yasin, S. F. M. (2010). 65–83. doi: 10.1080/13854046.2013.870231.
Validity of UKM1 Intelligence Test using rasch
analysis. Procedia-Social and Behavioral Sciences, 7, Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., &
205–209. doi: 10.1016/j.sbspro.2010.10.029. Tatham, R. L. (2014). Multivariate data analysis (Seven
Edit). Pearson Education Limited, NJ: Prentice hall.
Bergeron, R., & Floyd, R. G. (2013). Individual part
score profiles of children with intellectual disability: A Halberda, J., Mazzocco, M. M. M., & Feigenson, L.
descriptive analysis across three intelligence tests. (2008). Individual differences in non-verbal number
School Psychology Review, 42(1), 22 -38. acuity Correlate with Math Achievement. Nature, 455,
665–668. doi: 10.1038/nature07246.
Bond, T. G., & Fox, C. M. (2015). Applying the Rasch
model: Fundamental measurement in the human Hale, C. R., Casey, J. E., & Ricciardi, P. W. R. (2014).
sciences (Third Edit). New York, NY: Routledge. A cluster analytic study of the Wechsler Intelligence
Test for Children-IV in children referred for
Busato, V. V, Prins, F. J., Elshout, J. J., & Hamaker, C. psychoeducational assessment due to persistent academic
(2000). Intellectual ability, learning style, personality, difficulties. Archives of Clinical Neuropsychology: The
achievement motivation and academic success of Official Journal of the National Academy of
psychology students in higher education. Personality Neuropsychologists, 29(1), 75–85. doi: 10.1093/arclin/a
and Individual Differences, 29(6), 1057–1068. doi: ct086.
10.1016/S0191-8869(99)00253-6.
Buschkuehl, M., & Jaeggi, S. M. (2010). Improving Kim, S., Frisby, C., & Davison, M. (2004). Estimating
intelligence: A literature review. Swiss Medical Weekly, cognitive profiles using profile analysis via multi-
140(19), 266-272. doi: smw-12852. dimensional scaling (PAMS). Multivariate Behavioral

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

 The Intelligence Profile of the University Students 95

Research, 3171(May 2013), 37–41. doi: 10.1207/s1532 Morris, R., Blashfield, R., & Satz, P. (1981).
7906mbr3904. Neuropsychology and cluster analysis: Potentials and
problems. Journal of Clinical and Experimental
Komarraju, M., Ramsey, A., & Rinella, V. (2013). Neuropsychology, 3(1), 79–99. doi: 10.1080/01688638
Cognitive and non-cognitive predictors of college 108403115.
readiness and performance: Role of academic discipline.
Learning and Individual Differences, 24, 103–109. doi: Pelton, T. (2002). Where are the limits to the Rasch
10.1016/j.lindif.2012.12.007. advantage ? In Paper presented at the International
Objective Measurement Workshop (IOMW). New
Koski, L., Xie, H., & Finch, L. (2009). Measuring Orleans, LA.
cognition in a geriatric outpatient clinic: Rasch analysis
of the Montreal Cognitive Assessment. Journal of Primi, R. (2014). Developing a fluid intelligence scale
Geriatric Psychiatry and Neurology, 22(3), 151–160. through a combination of Rasch modeling and cognitive
doi: 10.1177/0891988709332944. psychology. Psychological Assessment, 26(3), 774-788.
doi: 10.1037/a0036712.
Landerl, K., Bevan, A., & Butterworth, B. (2004).
Developmental dyscalculia and basic numerical Rasch, G. (1960). Probabilistic models for some
capacities: A study of 8–9-year-old students. Cognition, intelligence and attainment tests. Danish Institute for
93(2), 99–125. doi: 10.1016/j.cognition.2003.11.004. Educational Research (Vol. 4). Copenhagen:
DanishInstitute for Educational Research. doi: 10.1016/
Landerl, K., & Moll, K. (2010). Comorbidity of learning S0019-9958(61)80061-2.
disorders: prevalence and familial transmission. Journal
of Child Psychology and Psychiatry, 51(3), 287–294. Rosander, P., Bäckström, M., & Stenberg, G. (2011).
doi: 10.1111/j.1469-7610.2009.02164.x. Personality traits and general intelligence as predictors
of academic performance: A structural equation
Lange, R. T., Iverson, G. L., Senior, G. J., & Chelune, modelling approach. Learning and Individual Differences,
G. J. (2002). A primer on cluster analysis applications to 21(5), 590–596. doi: 10.1016/j.lindif.201 1.04.004.
cognitive rehabilitation research. Journal of Cognitive
Rehabilitation, 20(1), 16–33. Retrieved from Schmidt, F. L., & Hunter, J. E. (1998). The validity and
http://psycnet.apa.org/record/2002-13993-002. utility of selection methods in personnel psychology:
Practical and theoretical implications of 85 years of
Lee, K., Ng, S.-F., Ng, E.-L., & Lim, Z.-Y. (2004). research findings. Psychological Bulletin, 124(2), 262-
Working memory and literacy as predictors of 274. doi: 10.1.1.172.1733.
performance on algebraic word problems. Journal of
Experimental Child Psychology, 89(2), 140–158. doi: Sideridis, G. D. (2011). The effects of local item
10.1016/j.jecp.2004.07.001. dependence on estimates of ability in the Rasch Model.
Rasch Measurement Transactions, 25(3), 1334–1336.
Letteri, C. A. (1980). Cognitive profile: Basic Retrieved from https://www.rasch.org/rmt/rmt253d.htm.
determinant of academic achievement. The Journal of
Educational Research, 73(4), 195–199. doi: 10.1080/ Spearman, C. (1904). “General intelligence,” Objec-
00220671.1980.10885234. tively determined and measured. The American Journal
Linacre, J. M. (2006). A user’s guide to of Psychology, 15(2), 201–292. Retrieved from
WINSTEPS/MINISTEP: Rasch-model computer https://www.jstor.org/stable/1412107?seq=1#metadata_i
programs. Retrieved from www.winsteps.com. nfo_tab_contents.

Mackintosh, N. J. (1998). IQ and human intelligence. Spearman, C. (1927). The abilities of man. New York,
Cambridge: Cambridge University Press. NY: Cambridge.

Mayes, S. D., & Calhoun, S. L. (2004). Similarities and Sumintono, B., & Widhiarso, W. (2015). Aplikasi
differences in Wechsler intelligence scale for children- pemodelan rasch pada assessment pendidikan (Rasch
Third edition (WISC-III) profiles: Support for subtest modeling application on educational assessment).
analysis in clinical referrals. Clinical Neuropsychologist, Cimahi: Trim Komunikata.
18(4), 559–572. doi: 10.1080/13854040490888530.
Tennant, A., & Conaghan, P. G. (2007). The Rasch
Mooi, E., & Sarstedt, M. (2011). A concise guide to measurement model in rheumatology: what is it and
market research: the process, data and methods using why use it? When should it be applied, and what should
IBM SPSS statistics. Heidelberg: Springer. one look for in a Rasch paper? Arthritis Care &
Research, 57(8), 1358–1362. doi: 10.1002/art.23108.

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1

96 Yudiana, Susanto, & Triwahyuni

Van der Ven, A., & Ellis, J. L. (2000). A Rasch analysis Woodcock, R. W. (1999). What can Rasch-Based scores
of Raven’s standard progressive matrices. Personality convey about a person’s test performance ? In S. E. E. &
and Individual Differences, 29(1), 45–64. doi: S. L. Hershberger (Ed.), The new rules of measurement:
10.1016/S0191-8869(99)00177-4. What every psychologist and educator should know (pp.
105–127). Mahwah, NJ: Erlbaum.
Warm, T. A. (1989). Weighted likelihood estimation of
ability in item response theory. Psychometrika, 54(3), Woodcock, R. W., McGrew, K. S., & Mather, N.
427–450. (2001). Woodcock-Johnson Tests of Cognitive Ability.
Itasca, IL: Riverside.
Weber, H. S., Lu, L., Shi, J., & Spinath, F. M. (2013).
The roles of cognitive and motivational predictors in Wright, B. D. (1996). Local dependency, correlations
explaining school achievement in elementary school. and principal components. Rasch Measurement
Learning and Individual Differences, 25, 85–92. doi: Transactions, 10(3), 509-511.
10.1016/j.lindif.2013.03.008.

Makara Hubs-Asia July 2019 | Vol. 23 | No. 1