5.4 Data collection
5.4.2 IMG questionnaire
5.4.2.2 Questionnaire validation
The Tasmanian IMG questionnaire had undergone face and construct validation between August 2011 and February 2012. Face validity was assessed in terms of the presentation and relevance of the questionnaire by myself and three academic staff members from various academic backgrounds. Construct validity was also assessed, to ensure the questionnaire measured what it set out to measure. There was some debate and concern about the use and wording of the original four point ‘forced choice’ Likert-scale format used by Hawthorne et al. (2003). It was felt the scales needed to reflect a five point Likert-scale, so as to allow neutrality or undecided choice to any posed question. However, the four point Likert-scale has been developed and is used at times to prevent the acquiescent response effect, where respondents answer the middle question at each question (Polgar, 1995).
Moreover, the wording used to determine the professional and non-professional satisfaction levels of IMGs in the original Likert-scales, was ‘very satisfactory’; ‘satisfactory’; ‘fair’; and ‘very unsatisfactory.’ Similarly, the wording used when determining the level of importance for certain aspects of residential location was ‘very important’; ‘important’; ‘not very important’ and ‘unimportant.’ It was felt the wording needed to be rectified to allow greater clarity of the rating scale currently used (Polgar, 1995). After much debate and discussion, the Likert-scales remain unchanged, so as to make more accurate parallel comparison between the Victorian and Tasmanian study results. Despite this concern being addressed, there were still a number of concerns regarding the four point Likert-scale and wording of the scales from the academic staff and myself. These concerns became more apparent as the results of the questionnaire were analysed.
In addition to these initial concerns, other psychometric properties of the three principal Likert-scales in the questionnaire were evaluated before the data were analysed. These Likert-scales focused on an IMG’s satisfaction with the aspects of
their current professional position and non-professional lifestyle. The 43 items from the three Likert-scales were subjected to Principal Components Analysis (PCA) with a Varimax rotation using SPSS 20.0. This analysis was conducted to extract the maximum amount of variance across each of the three Likert-scales (Pallant, 2011). Data suitability was assessed prior to performing PCA where items were excluded if loading of coefficients were less than 0.5 (Pallant, 2011). In addition, the Kaiser- Meyer-Oklin measure of sampling adequacy was shown to be .666, which is above the recommended value of .6 and Bartlett's Test of Sphericity was .000, which is statistically significance, and supports the factorability of the correlation matrix (Pallant, 2011).
To identify and label each component, an additional exploration of the highest loaded items was undertaken. PCA revealed 12 components were shown to have Eigen values above 1, as shown in Figure: 5.3.
Figure: 5.3 Principal Components Analysis scree plot
However, when parallel analysis was conducted using Monte Carlo PCA for Parallel Analysis (Watkins, 2000), the results showed only four components with a total of
32 questions were suitable to be retained. These four components explained
24.85%, 12.07%, 6.79% and 5.08% of the variance respectively as presented in Table 5.1. The four components were identified and labelled ‘the importance of
satisfaction with employment and community’; ‘the perceived socio-cultural factors influencing future employment’; ‘the factors which improve satisfaction of the quality of life’; and ‘the perceived workplace factors influencing future
Table 5.1: Principal Components Analysis results for Likert-scale questions
Varimax rotation Component
1 2 3 4
Q37a. Level of satisfaction with type of work
Q38c. Non-professional aspects of friendliness of people
Q37k. Level of satisfaction with friendliness of the local community Q37d. Level of satisfaction with good/ supportive colleagues Q37j. Level of satisfaction with friendliness of your patients Q38a. Non-professional aspects of appeal of location
Q37f. Level of satisfaction with level of professional support Q37h. Level of satisfaction with medical facilities/resources Q37e. Level of satisfaction with salary level
Q38i. Non-professional aspects of access to cultural community and resources
Q37b. Level of satisfaction with medical location
Q37c. Level of satisfaction with relevance to your skills/past experience
Q38m. Non-professional aspects of access to private transport (own car)
Q37g. Level of satisfaction with access to training/ supervision Q38b. Non-professional aspects of size of city/town
Q38j. Non-professional aspects of access to social activities
.777 .768 .755 .706 .682 .679 .653 .632 .626 .617 .615 .614 .580 .569 .534 .532 Component 1 2 3 4
Q39m. Influence future work: Access to cultural community and resources
Q39o. Influence future work: Access cultural or religious foods or goods
Q39n. Influence future work: Access to social activities Q39k. Influence future work: Access to religious facilities
Q39b. Influence future work: Improved medical facilities/resources Q39p. Influence future work: Access to public transport
Q39l. Influence future work: Access to friends/ family members Q39h. Influence future work: Settlement near cultural community
.760 .738 .681 .644 .643 .643 .599 .598 Q38l. Non-professional aspects of access to public transport
Q37l. Level of satisfaction with access to public transport
Q38k. Non-professional aspects of access cultural or religious foods or goods
Q38e. Non-professional aspects of access to employment for partner/ spouse
Q38d. Non-professional aspects of quality of facilities (transport, shops etc.)
Q38f. Non-professional aspects of access to good schools
.859 .747 .670 .647 .644 .540 Q39d. Influence future work: Shorter working hours
Q39e. Influence future work: Less on-call responsibility
.781 .664 Percentage (%) of variance explained 24.85 12.07 6.79 5.08
In addition to factor analysis, internal consistency was used to assess reliability of each of the components prior to any subsequent analysis. Cronbach Alpha is the most commonly used method of assessing reliability. Thus, Cronbach Alpha coefficient of a scale above 0.7 demonstrates an acceptable level of internal reliability, while scores lower than 0.7 indicates an unacceptable level of internal reliability (Munro, 2005; Pallant, 2011). The results of the internal reliability of the four components indicate a high degree of reliability as presented in Table 5.2.
Table 5.2: Cronbach’s Alpha reliability coefficients of factors