Padjadjaran Interest Inventory : Evaluation of Psychometric Properties

This research aims to evaluate the psychometric properties of a new instrument for measuring vocational interest: Padjadjaran Interest Inventory (PII). There were 2,648 participants in this study, consisting of workers, high school, and university students, with gender proportion of 1,014 (38.3%) males and 1,634 (61.7%) females. This research used descriptive statistic test, t-test, and MANOVA for gender differences, reliability coefficients and validity evidence by using confirmatory factor analysis (CFA). The results showed that PII has a good psychometric properties: it has good reliability, and valid internal structure; it is standardized by gender; and it is applicable for large groups with relative ease. PII can be used for career exploration. Limitation of this study was discussed for future research.

Career 1 is a continuous process occurring throughout an individual's lifetime (Sharf, 2006).In deciding a future career, vocational interest is one of the important factors which helps counselors direct an individual (Betz, Harmon, & Borgen, 1996).Individual interest in certain areas in the education process impacts attention, goals, and levels of learning (Hidi, 1990).In addition, interest also affects a person's achievement in the career area (Jansen, Lüdtke, & Schroeders, 2016;Li & Yang, 2016) and a person's success in education area (Bloye, 2007).
In Indonesia, however, instruments used to measure vocational interest are still limited, i.e., Kuder Preference Record Form C (Kuder, 1948) and The Rothwell Miller Interest Blank (RMIB) (Miller, 1958;Rothwell, 1947).Kuder Preference Record 1 Address for corespondence: whisnu.yudiana@unpad.ac.id; ilham1301@mail.unpad.ac.id; Form C basically has four weaknesses (Kelly, 2002), i.e., the choices of activity and occupation are irrelevant to the times; it is limited in paper-based test; it is not linked to Holland's six structures (Holland, 1997), which is the theory that underlines vocational interest; and it does not show profiles of interests.
On the other hand, the RMIB test also has weaknesses in the aspect of the instrument's reliability and validity (Yudiana, 2011).In RMIB, only scientific interest area is reliable, four areas are quite reliable, and eight areas are unreliable.In addition, based on confirmatory factor analysis, this model is significantly different from the construct being measured.Therefore, it can be said that this measure has not been tested for validity.
Therefore, using this instrument may lead to mismeasurement.This situation illustrates that Indonesia lacks an interest inventory that is relevant to the present times and today's context.Paradigm shifts have caused major changes in the world of work (Kuhn, 1962).This further increases the importance of innovative approaches to career counseling (Maree & Taylor, 2016).Due to its importance in career counseling (Sharf, 2006), development of interest inventory that is relevant to the times and today's paradigm becomes important.The most recently developed interest inventory that can cover the weaknesses of Kuder Preference Form C and RMIB is Padjadjaran Interest Inventory (PII) (Yudiana, Purwono, & Wiyono, 2011).
PII measures three aspects in 18 interest areas which are divided into eight basic interests, five high interests, and five low interests.In addition, it has been reported that PII has good psychometric property.PII has a reliability coefficient of between 0.794 -0.934 in all dimensions and has validity evidence based on test content, response processes, and internal structure (Yudiana et al., 2011).
However, there are several limitations to previous studies (Yudiana et al., 2011).Firstly, the norming used in PII was last reviewed in 2011.Whereas, updating the norming of an instrument is vital in order that the instrument remains reliable and valid (Suwartono, 2016).Secondly, the number of participants involved in the norming was limited to 550 individuals.Thirdly, there has never been any report on score differences between male and female, whereas gender plays a role in career choices that make norming be differentiated between genders (Elena, 2014;Lawson, Lee, Crouter, & Mchale, 2018;Volodina & Nagy, 2016).Fourthly, the age range of career stages used in this norming was limited between 17 and 19 years old, which falls into only one career stage, namely exploration (Super, 1990).Based on these, this research is an improvement of previous researches to respond to those limitations.This research aims to determine the reliability of PII and collect its internal validity evidence with larger target participants, not only in terms of age range but also career stages of the participants, i.e., from exploration to higher level (Super, 1990).In addition, it aims to view the comparison of scores between male and female.

Instrument
The instrument used in this research was the Padjadjaran Interest Inventory (PII) (Yudiana et al., 2011).PII consists of two item formats, i.e., activity with two types of response options: preference and competence belief, and occupation.There are 144 items for each item format, of both activity and occupation.Inactivity item format, participants were asked to scale twice; the first is their preferences for the activity (1 = strongly dislike and 7 = strongly like) and the second is their competence belief to do the activity (1 = unable to do and 7 very competent).On the other hand, in occupational item format, participants were asked to scale their occupational preferences (1= strongly dislike and 7 = strongly like).The format choices are based on Personal Globe Inventory (PGI) test (Tracey, 2002) that measure the same concept.PII describes 18 interest areas in three item formats thus, there are total 54 score scales reported by PII.Each scale is measured by eight items.Eighteen interest areas are explained in further detail in Table 2.

Analysis
Descriptive statistics (mean, standard deviation, skewness, and kurtosis) for 54 interest area scales were calculated to describe details of samples in this research.The t-test for independent samples was used to determine the differences in interest area scales by gender.For testing overall gender differences in interest areas, multivariate analysis of variance (MANOVA) was used.The technique used to determine the reliability of 54 interest area scales was Cronbach's alpha based on reliability coefficient obtained by calculating the standard error of measurement (Kaplan & Sacuzzo, 2005).
The validity evidence in this research was based on internal structure.Validity evidence based on internal structure refers to dimensionality or underlying factors measured in an instrument (Sireci & Sukin, 2013).The confirmatory factor analysis (CFA) was used to achieve this validity.In this research, the hypothesized model is that every interest area is measured by two item formats, i.e., activity and occupation, and each item format is measured by eight items.Considering that, inactivity item format, there are two types of response options, i.e., preference and competence belief, the hypothesis of a measurement model for each interest area is tested by second-order factor analysis, which is shown in detail in figure 1.
There were eighteen CFA tests describing eight interest areas.Indicators for the goodness of fit in determining model fit are CFI, GFI, NFI and NNFI (McDonald & Ho, 2002).The fitness criteria are that RMSEA value of less than 0.08 indicates reasonable fit, and less than 0.05 indicates a very good fit.Then, values of CFI, GFI, and NFT of more than 0.90 is considered as satisfactory model fit.On the other hand, NNFI value is no less than 0.80 (Hooper, Coughan, & Mullen, 2008).Descriptive statistical analysis was conducted by using Statistical Product and Service Solution (SPSS) 22.0 for Windows.On the other hand, the reliability was tested using the psych package (Revelle, 2017) in R and CFA programming using lavaan package on R programming (Rosseel, 2012).

Procedure
Data collection was conducted in the BPIP and Psychological Bureau in Bogor, based on the data collection process in the psychological test for the purpose of recruitment and employee selection between 2011 to 2017.PII was used as one of the instruments in the psychological test.The same procedure was used in data from TPBK, namely deriving from a psychological test for the mapping of student potential in the Faculty of Psychology, Universitas Padjadjaran between 2013 to 2017.On the other hand, data from the research was sourced from eight high schools in Bandung.Administration Process was conducted according to the manual (Yudiana et al., 2011) by final-year students of Psychology who already had test method subject.

Descriptive Statistics
This section elaborates mean score, standard deviation, skewness and kurtosis of PII instrument, as shown in Table 3.The minimum score of each area in PII is 8 and the maximum score is 56.In Activity item format, ManWo had the lowest mean (20.04) and Help had the highest mean (38.82).In Competence item format, ManWo had the lowest mean (18.89) and Help had the highest mean (34.49).In Occupational item format, ManWo had the lowest mean (17.30) and Man had the highest mean (35.23).In all three item formats, there isn't any skewness greater than 2 or less than -2.The same applies for kurtosis, there isn't any area of the three item formats greater than 2 or less than -2.This shows the normality of data (Field, 2009).

Reliability
Cronbach's alpha was used in 18 interest areas stated in Table 4.It can be seen that estimated reliability for all interest areas is relatively high with rvalue > 0.70, with the lowest score is Man area in Occupation item format (0.81) and the highest score is BusDe (0.92) in Activity item form.It shows that contents in item formats are homogeneous.

Validity evidence based on the internal structure
Validity evidence by second-order confirmatory factor analysis (CFA) showed that the data fit the hypothesized model.This is based on Table 6 presents a summary of factor loading for each interest area.In almost all interest areas, there are items with a loading factor < 0.5.The number of items for the activity, competence, and occupational item formats is 16, 14, and 11, respectively (total is 41 items).Interest areas with the highest number of lowvalue items were DatPr and Art, with seven items each.However, the value of Average Variance Extracted (AVE) for each interest area in the second stage of CFA was > 0,5, with interest area with the lowest value was ManWo (0.64) and interest area with the highest value was SosSc (0.79).And, for all interest areas, the value of construct reliability was higher than the standard set, which is > 0.8.

Discussion
This research aims to evaluate the psychometric properties of PII.The results show that PII had a good psychometric property.Therefore, it is expected that PII can be utilized and can contribute to the process of measuring interest that is relevant to the context in Indonesia.In addition, it is expected that PII can be used for purposes such as career counseling or the search for vocational interest itself.Results of descriptive statistics indicate that several interest areas had a higher value than other areas.This is indeed in line with the estimation that vocational interests are not evenly distributed (Maree & Taylor, 2016;Tracey, 2002).Moreover, based on the spherical model of interest, interests are divided into basic interests, high prestige interests and low prestige interests which certainly explain why several areas are more preferred than others (Tracey, 2002).
Estimated reliability value of all item formats and interest areas in PII also show relatively high value.Similar results were also indicated in other researches based on a spherical model of interest (Etzel et al., 2016;Long, Adams, & Tracey, 2005;Šverko, 2008).Thus, items in each interest area together consistently measured the same construct.It still, however, requires further research on the reliability evidence by repeated tests to determine the stability of measurement.
Other result shows good evidence on the internal structure of PII.In the analysis using secondary confirmatory factor analysis, almost all model fit indices met the standard set (Hooper et al., 2008), meaning that the data fit the hypothesized model.Therefore, this can assist in providing evidence that the score resulted from PII can be used as a basis for interpreting the score produced.However, it still requires further examination on items with factor loading not meeting the standard.
The difference in interest preferences between gender is indeed in accordance with previous researches (Tracey, 2002;Elena, 2014;Lawson et al., 2018;Volodina & Nagy, 2016).Male participants had more of a preference on activities, felt more capable and also had more of a preference on occupations in Man, DatPr, Mec, BusSys, FinAn, QuaC, ManWo, and ConsRep.Male participants also had more of a preference on occupations in Sci, but not in Man.Meanwhile, female participants had more of a preference on activities, felt more capable, and had more of a preference on occupations in SosF, Art, Help, Scie, SocSc, PerSer, and BaSer; but felt less capable in BaSer and had less preference on occupations in Scie.The significant difference of scores in those interest areas serves as a basis in the norming of PII which is differentiated between male and female, and so a value of the norm will be equivalent.
Although the empirical results indicate good reliability value and validity evidence of PII, there are several limitations in this research that require consideration.First, despite that the sample size was quite large and varied by age, gender and employment status, they were geographically limited as most of the data was only from one region, namely West Java and a small section of Jakarta.The l`arge sample size that represents Indonesia is essential, considering that Indonesia varies in terms of ethnicity in each region.In addition, the sample size for the employed status was relatively limited, and therefore this could be expanded.
The second limitation is related to validity evidence.This research still focused on validity evidence based on internal structure.PII resulted significantly higher scores compared to other existing interest and personality instruments.Therefore, further research regarding relationship between scores obtained and similar instruments is required.This can help facilitate the interpretation of scores resulted from PII.
The third limitation is related to the considerable number of items and the existence of relatively low factor loading items.Development of PII with a more briefer format without compromising quality of the result is essential for the time efficiency of data collection.Research on this briefer format corresponds with the researches on existing interest inventories (Tracey, 2010).Item response theory approach can be used to rule out the low factor loading items.

Conclusion
PII is a new interest inventory that can help identify individual interest in future career direction.The advantages of this instrument are that it doesn't only measure individual vocational interest, but also individual competence belief to perform the career choice.This certainly could assist individuals with unclear career choices.PII has good psychometric properties, both in terms of consistency and the measurement fit with the construct being used.In addition, this research finds that there are gender differences in interest preferences that the norming by gender becomes important.This research also proves that PII can be used for individuals within Super's career stage of exploration (Super, 1990).Therefore, PII can help individuals in career choices and assist the process of career counseling in order to make the right career choice.

Recommendation
Further research on the psychometric property should be directed in the expansion of sample in relation to norms in Indonesia, predicted validity and other instruments, and the development of PII with a briefer format using item response theory.In addition, due to the advancement of technology, development of web-based PII should be considered as it faces a generation intensely engaging with technology.Therefore, PII can have promising development in measuring vocational interests.interest blank (Miller revision).Melbourne: Australian Council for Educational Research.

Table 2 .
Description of interest area 12 Financial Analysis focuses on helping others with financial issues FinAn 13 Science focuses on general interest in science Scie 14 Quality Control focuses on checking details QuaC 15 Manual Work focuses on working with hands or simple machines ManWo 16 Personal Service focuses on working with people in everyday transactions PerSer 17 Construction & Repair focuses on working with machinery to repair and build ConsRep 18 Basic Services focuses on selling products and services Information BasSer Figure 1.Measurement Model for Every Interest Area

Table 5 ,
showing indices of the goodness of fit.In more detail, RMSEA falls within a range of 0.05 -0.08.Almost all values for CFI, GFI, NFI, and NNFI are > 0.90, except that in DatPr and SosSc the GFI values are 0.88.

Table 6 .
Summary of factor loading, Average Variance Extracted (AVE) and construct reliability