Professional Documents
Culture Documents
/ Cohort 8
The data in this report was collected from international students studying at a four-year regional
university in Sothern California. The purpose was to understand the strength of some factors that
motivated these students to travel and study at the campus where the data was collected. The instrument
was pen-and-paper questionnaire consisting of ten five-Likert scale type items ranging from strongly
disagree, coded as one, to strongly agree, coded as five. The ten items in the instrument examined the
strength of factors constituting international students college choice construct. Five of these items were
pull factors: (a) Rounding of Education through International Study, (b) Personal Fulfilment, (c) Possibility
to Emigrate, (d) Possibility to Establish International Connections, and (e) Availability of Scholarships.
There were also five push factors: (a) Unavailability of Similar Program at Home Campuses, (b) Difficulty
to Be Accepted at a Good Program at Home, (c) Desire to Be Away from Home, (d) Family Pressure, and
(e) Peer Pressure.
Minimum
Maximum
Mean
Median
Mode
Std. Deviation
Round Edu.
50
3.60
4.00
4.00
1.1066
No Similar at Home
50
2.96
3.00
2.00
1.2282
Difficult at Home
50
3.34
3.50
4.00
1.0994
Personal Fulfilment
50
3.74
4.00
4.00
1.1031
Emigration
50
3.38
3.00
4.00
1.1045
International Connections
50
2.96
3.00
2.00
1.2282
50
3.38
4.00
4.00
1.0476
Family Pressure
50
2.88
3.00
2.00
1.3346
Peer Pressure
50
3.14
3.00
4.00
1.1068
Scholarships
50
2.84
3.00
3.00
1.1132
Table 1 above illustrates the mean, median, modes, range and standard deviations for subjects
responses on the ten items within the factors assessed. The table shows that subjects responses ranged
from one to five on all variables, and means varied with Personal Fulfilment as the factor with the highest
score followed by Rounding Education through International Study, indicating that these were the most
Page 2 of 10
influential choice factors for the subjects in this report. On the other hand, the Availability of Scholarships
was the least influential factor, followed by Family Pressure to Study Abroad. These variations are also
demonstrated in Table 2 below through item-difficulty-to-embrace scores. These were calculated for each
item by summing the responses of subjects to each different item then dividing that by the product of the
number of subject and 5, which is the number of points in the Likert scale employed. In Table 2 below,
scores are of one with higher values indicating variables that were more likely to be embrace by subjects.
Table [2]: Factors Affecting International Students College Choice Item Difficulty & Varp
Round
Edu.
Item
Item
Difficulty
No Similar
at Home
Difficult at
Home
Personal
Fulfilment
Emigration
Internation
al Conn.
Away from
home
Family
Pressure
Peer
Pressure
Scholarshi
ps
0.72
0.59
0.67
0.75
0.68
0.59
0.68
0.58
0.63
0.57
Varp
1.2
1.48
1.18
1.19
1.20
1.48
1.08
1.75
1.20
1.21
Item total
180
148
167
187
169
148
169
144
157
142
In addition to item difficulty, Table 2 above reports on the variance for the items which is used to
calculate the Reliability Coefficient which produced a value of 0.41 (rxx = 0.41). This suggests that the data
collected for this report may not be sufficiently reliable for generalization to the population. This may be
attributed to the small number of subjects who participated in the report or more likely to the small
number of items in the instrument, and suggests that a bigger sample size and/or an instrument with
more items may need to be employed for relevant claims about factors affecting international students
college choice to be made.
On the other hand, the total x varp was 21.61 and the standard deviation of scores was 4.65.
(Total X Varp = 21.61; SD = 4.65). The value of standard deviation was then used together with the
reliability coefficient to calculate the Standard Error of Measurement producing a value of 3.58 (SEM =
3.58). This value for standard deviation of errors of measurement demonstrated the extent to which the
data provide accurate scores.
Further analysis of the data following Bond and Fox (2007) described methodology for the Rasch
model is done with WINSTEPS software Version 3.68.2 (Linacre, 2010). Analysis shows that the data are
good fit for the model as demonstrated in the variable map in Figure 1 below and discussed afterwards.
Page 3 of 10
Scholarships
No Similar
Difficult
Emigration
Figure 1 above shows that the analysis resulted in a scale consisting of items that range between
one logit for item difficulty and two logits for person ability. The mean for the item calibrations is slightly
below that for the subjects. Four of the items fall between the mean and one standard deviation, two
items at one standard deviation and two items between one and two standard deviations. As for subjects,
most of these fall within one standard deviation for subjects. Only three subjects were at or above two
standard deviations.
Page 4 of 10
Overall, subjects were distributed with more intensity closer to the mean for subjects and items,
with fewer subjects to the end of the logit scale. Figure 1, however, suggests that there is some level of
imprecision. There are two subjects who are below items, and eleven subjects who are above the most
difficult item in the instrument. This indicates that the instrument can be improved or modified, by
including items at the upper end of the scale that would be targeted at the level of the persons at the
upper end of the scale.
WINSTEPS was further employed to calculate INFIT and OUTFIT Mean Squares with an expected
value of 1, and 1.5 and 0.70 as arbitrary cut off points to determine misfitting responses; and Standardized
Z with an expected value of 0 and arbitrary cut points of 2.0 and -2.0. The summary statistics of the data
are in Table 3 below.
TABLE [3]: Summary Statistics for Factors Affecting International Students Choice
INPUT: 50 Internationals 10 Factors MEASURED: 50 Internationals 10 Factors 5 CATS
-------------------------------------------------------------------------------SUMMARY OF 50 MEASURED Internationals
------------------------------------------------------------------------------|
RAW
MODEL
INFIT
OUTFIT
|
|
SCORE
COUNT
MEASURE
ERROR
MNSQ
ZSTD
MNSQ
ZSTD |
|-----------------------------------------------------------------------------|
| MEAN
32.2
10.0
.18
.31
1.00
.0
1.00
.0 |
| S.D.
4.6
.0
.44
.03
.43
1.2
.42
1.1 |
| MAX.
42.0
10.0
1.23
.40
2.37
2.9
2.27
2.7 |
| MIN.
22.0
10.0
-.75
.29
.21
-2.9
.19
-3.0 |
|-----------------------------------------------------------------------------|
| REAL RMSE
.33 ADJ.SD
.29 SEPARATION
.86 Intern RELIABILITY .43 |
|MODEL RMSE
.31 ADJ.SD
.31 SEPARATION 1.02 Intern RELIABILITY .51 |
| S.E. OF International MEAN = .06
|
------------------------------------------------------------------------------International RAW SCORE-TO-MEASURE CORRELATION = 1.00
CRONBACH ALPHA (KR-20) International RAW SCORE RELIABILITY = .44
SUMMARY OF 10 MEASURED Factors
------------------------------------------------------------------------------|
RAW
MODEL
INFIT
OUTFIT
|
|
SCORE
COUNT
MEASURE
ERROR
MNSQ
ZSTD
MNSQ
ZSTD |
|-----------------------------------------------------------------------------|
| MEAN
161.1
50.0
.00
.14
1.00
.0
1.00
.0 |
| S.D.
14.8
.0
.28
.00
.17
1.0
.17
.9 |
| MAX.
187.0
50.0
.35
.15
1.40
2.2
1.39
2.2 |
| MIN.
142.0
50.0
-.50
.13
.84
-1.0
.85
-.9 |
|-----------------------------------------------------------------------------|
| REAL RMSE
.14 ADJ.SD
.24 SEPARATION 1.69 Factor RELIABILITY .74 |
|MODEL RMSE
.14 ADJ.SD
.24 SEPARATION 1.76 Factor RELIABILITY .76 |
| S.E. OF Factor MEAN = .09
|
------------------------------------------------------------------------------UMEAN=.000 USCALE=1.000
Factor RAW SCORE-TO-MEASURE CORRELATION = -1.00
500 DATA POINTS. LOG-LIKELIHOOD CHI-SQUARE: 1389.99 with 438 d.f. p=.0000
Page 5 of 10
Table 3 above shows the mean for the raw score is 32.2 with a standard deviation of 4.6 (M =
32.2; SD = 4.6). The maximum score is 42.0, and the minimum score is of 22.0. The Measure calculation
and Model Error are respectively .18 and .31 for the mean, .44 and .03 for the standard deviation, 1.23
and .40 for the maximum score, and -.75 and .29 for the minimum score. The analysis suggests that the
data fits the Rach model very well since MNSQ produced a value of 1.00 and ZSTD produced a value of .0
for INFIT and OUTFIT measures. However, and as demonstrated in traditional statistical analysis above,
the internal reliability is quite low at .43 and Cronbach Alpha is at .44. This may be attributed to the items
in the instrument as the scelogram in Figure 1 above suggested that items are not well spread along the
scale.
Further investigation of the data fit for the model is done through fit order analysis for the
individual items in the questionnaire. Table 4 below illustrates some descriptive statistics together with
INFIT/OUTFIT MSQ and ZSTD, as well as PT-Measure correlation and expected values and expected match
for each individual item in the instrument. The table below puts items in their misfit order.
TABLE [4]: Individual Item Fit for Factors Affecting International Students Choice
INPUT: 50 Internationals 10 Factors MEASURED: 50 Internationals 10 Factors 5 CATS
-------------------------------------------------------------------------------International: REAL SEP.: .86 REL.: .43 ... Factor: REAL SEP.: 1.69 REL.: .74
Factor STATISTICS:
MISFIT ORDER
-------------------------------------------------------------------------------------------------|ENTRY
TOTAL
MODEL|
INFIT | OUTFIT |PT-MEASURE |EXACT MATCH|
|
|NUMBER SCORE COUNT MEASURE S.E. |MNSQ ZSTD|MNSQ ZSTD|CORR. EXP.| OBS% EXP%| Factor
|
|------------------------------------+----------+----------+-----------+-----------+-------------|
|
8
144
50
.31
.13|1.40
2.2|1.39
2.2|A .31
.42| 18.0 31.6| Family
|
|
10
142
50
.35
.13|1.18
1.1|1.17
1.0|B .12
.42| 22.0 31.7| Scholarships|
|
4
187
50
-.50
.15|1.08
.5|1.08
.5|C .37
.37| 34.0 39.7| Personal
|
|
3
167
50
-.10
.14| .99
.0| .98
-.1|D .34
.41| 34.0 32.6| Difficult
|
|
1
180
50
-.35
.14| .97
-.1| .93
-.3|E .46
.39| 40.0 37.3| Round Edu. |
|
9
157
50
.08
.13| .94
-.3| .95
-.2|e .36
.42| 22.0 31.2| Peer
|
|
7
169
50
-.14
.14| .94
-.3| .91
-.5|d .31
.40| 36.0 32.6| Away
|
|
5
169
50
-.14
.14| .86
-.8| .85
-.9|c .50
.40| 38.0 32.6| Emigration |
|
2
148
50
.24
.13| .84 -1.0| .85
-.9|b .64
.42| 26.0 31.4| No Similar |
|
6
148
50
.24
.13| .84 -1.0| .85
-.9|a .64
.42| 26.0 31.4| Connections |
|------------------------------------+----------+----------+-----------+-----------+-------------|
| MEAN
161.1
50.0
.00
.14|1.00
.0|1.00
.0|
| 29.6 33.2|
|
| S.D.
14.8
.0
.28
.00| .17
1.0| .17
.9|
| 7.3
2.8|
|
--------------------------------------------------------------------------------------------------
As demonstrated in Table 4 all items fitted the model with Family Pressure and Scholarships items
as the closest to the cut off point of 1.50 for MNSQ (IFIT MNSQ = 1.40, OUTFIT MNSQ = 1.39; IFIT MNSQ
= 1.18, OUTFIT MNSQ = 1.17 respectively) . Furthermore, Family item exceeded the cutoff point of 2.00
for ZSTD (INFIT ZSTD = 2.2, OUTFIT ZSTD = 2.2). On the other hand, PT-MEASURE produced positive values
for all items with small deviation between the observed correlation and expected value indicating that
data was consistent with the model expected scores. A deeper examination of the data is illustrated in
Table 5 and Table 6 below.
Page 6 of 10
Page 7 of 10
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Strongly Disagree
Disagree
Neutral
Agree
Strongly Agree
Table 5 above reports on the frequencies of five different options of the five-Likert scale used in
the instrument in relationship to expected responses. The least fitting items are at the top of the table,
and some categories in the table are marked with asterisks, which indicate that the mean for the lower
category is higher than the mean of persons who responded by choosing a higher category. In other
words, while we expect subjects with a higher mean logit measure to select the categories with higher
values than those with lower values, this was not the case in the categories marked with asterisks. These
categories where average measure did not ascend with category score were Agree and Strongly Agree for
Family item, Disagree and Strongly Agree for Scholarship item, Agree and Strongly Agree for Personal
Fulfilment item, Disagree for Peer Pressure item, Strongly Agree for Being Away from Family item, and
Neutral for Possibility to Emigrate item.
Finally, Table 6 above identifies subjects who had unexpected responses for the items in the scale.
There were a total of 12 such subjects out of the total sample of 50 in this report. For example, subjects
35 and 40 marked the item Family as 5 while they were expected to mark that item towards the lower
end of the scale. One subject gave three unexpected responses, subject 45, on Personal, Difficult and
Away items. This suggests that the data and report conclusions will benefit from excluding this particular
subjects responses from the analysis.
Page 8 of 10
Page 9 of 10
References
Bond, T. & Fox, C. (2007). Applying the Rasch model: Fundamental measurement in the human sciences.
Lawrence Erlbaum.
Hemsley-Brown, J. V. & Oplatka, I. (2006). Universities in a competitive global marketplace: A systematic
review of the literature on higher education marketing. International Journal of Public Sector
Management, 19 (4), 316 338.
Lee, E. S. (1966). A theory of migration. Demography, 3(1), 47 57.
McMahon, M. E. (1992). Higher education in a world market: An historical look at the global context of
international study. Higher Education, 24, 465 482.
Page 10 of 10