Page 64 - Pengurusan Prestasi Nilai Teras POISE UM

Rating scale research is important to check the possibility of scale
modifications that was done (Bond and Fox, 2015. A review of
the rating scales used in the NTP instruments (Very unsatisfactory,
Unsatisfactory, Somewhat Satisfactory, Satisfactory, Very Satisfactory)
was carried out by performing a scale calibration which is unique to
the Rasch Measurement Model. From the analysis, empirical data
showing how respondents maximise each scale in the instrument
could be identified, thus the ability and understanding of respondents
in differentiating each scale in the NTP instrument could be examined.
There are six review criteria for the functionality of the rating scale
(Kreijns et al., 2020), namely:
1. There are at least 10 observations for each category
2. The average value of observations must increase monotonically
(uniformly)
3. The threshold value must increase monotonically as the
categories increase
4. The outfit MNSQ value is less than 2.00
5. The distances between categories and adjacent categories are
between 1.4 logit and 5.00 logit
64 6. The category probability curve shows a significant peak

Table 5.4: Rating scale structure

Category Observation Average Outfit Andrich
scale frequency value of MNSQ threshold
observation
2 5 - 2.69 0.78 NONE
3 61 0.34 1.39 - 4.04
4 182 2.35 0.84 - 0.17
5 347 5.65 0.87 4.21
Findings show that only scales 2, 3, 4 and 5 were selected by the
respondents and there was no frequency for scale 1. The average
observation of each category was found to increase uniformly with
the category scale. The outfit MNSQ value for each category is also
less than 2.00. Moreover, the distance between categories (Andrich
threshold) remains in the range of 3.87 logit to 4.38 logit (See Table
5.5). The increase in distance value between categories that were good
could be seen through the clear peak shape in Figure 5.1. Based on
Table 5.4, the functionality of the NTP instrument rating scale has met
five of the six criteria listed.