PHASE ONE RESPONDENTS
5.4 Selective EPQ Code Validation
5.4.2 Testing the Components of Perceived EPQ
The OEPQ Rating is important in developing a reliable and valid measure of Perceived EPQ. As noted in section 4.5.5, having a separate and independent measure of the construct allows an initial assessment of how well the proposed EPQ Scale measures Perceived EPQ. A close relationship between two independent measures (OEPQ Rating and EPQ Scale) provides an indication that one is accurately measuring the phenomena of interest (Parasuraman et al. 1988). Where no construct measure exists, this ‘dummy criterion’ method is a popular approach to in scale development and validation (Parasuraman et al. 1988; Pitt et al. 1995; Reynoso & Moores, 1995; Finn et al. 1996).
Figure 25 shows a scatter plot of average EPQ variable scores (Selective codes) against OEPQ Rating scores. The plot indicates a strong positive relationship between the two sets of figures and further evidence of this is provided in table 27. The strong and significant correlation between the two independent variables gives an early indication of validity for the proposed set of Perceived EPQ components and suggests that no critical variables have been omitted.
Figure 25. Scatter plot: Average EPQ Variable to OEPQ Rating
8.00 6.00 4.00 OEPQ Rating (1-9) 7.50 7.00 6.50 6.00 5.50 5.00 4.50 Av erag e E PQ V ari abl e R Sq Linear = 0.757
Exploring Perceived EPQ. Alistair Brandon-Jones
Table 27. Pearson Correlation: OEPQ Rating & Average EPQ Variable
OEPQ Rating (1-9) Average EPQ Pearson Correlation .870(**) Sig. (2-tailed) .000 N 35
** Correlation is significant at the 0.01 level (2-tailed).
Additionally, the proposed componentsare validated by examining the power of scale scores to predict the independent criterion (Flynn et al. 1990). Linear regression was carried out to assess the extent to which the average EPQ score predicts variance in OEPQ Rating (Table 28). For the 35 phase two interviewees, 75% of variance in OEPQ was explained by the average EPQ score.
Table 28. Linear Regression: Average EPQ to OEPQ Rating
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .870(a) .757 .750 .59164
a Predictors: (Constant), Average EPQ
Whilst the average EPQ score is a useful ‘yardstick’ for prediction, a combination of variables may be even more powerful in predicting OEPQ Ratings. Therefore regression based on all variables is carried out (Table 29), showing an adjusted R2 of .913. This indicates that a combination of all variable scores predict 91.3% of variance in OEPQ Ratings.
Table 29. Linear Regression: All EPQ Variables to OEPQ Rating
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .994(a) .987 .913 .34912
a Predictors: (Constant), All 29 EPQ Variables based on phase one interviews
However, it is clear that some variables will have more influence than others in explaining OEPQ. Step-wise regression means the variable contributing most to prediction is added first, with others added based on their incremental improvement to the model. Table 30 shows the best solution with five variables (system navigation, content, problem resolution, on-time delivery, and FMS integration) that explain 92.7% of variance in OEPQ Rating scores. The adjusted R2 is even higher than with all variables and reflects the value of model parsimony in prediction.
Exploring Perceived EPQ. Alistair Brandon-Jones
Table 30. Stepwise Regression: All EPQ Variables to OEPQ Rating
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .781(a) .609 .598 .75062 2 .858(b) .737 .720 .62601 3 .912(c) .832 .816 .50809 4 .941(d) .886 .871 .42552 5 .941(e) .886 .875 .41876 6 .959(f) .919 .908 .35843 7 .968(g) .937 .927 .32039
a Predictors: (Constant), Attitudes
b Predictors: (Constant), Attitudes, System Navigation c Predictors: (Constant), Attitudes, System Navigation, Content
d Predictors: (Constant), Attitudes, System Navigation, Content, Problem Resolution e Predictors: (Constant), System Navigation, Content, Problem Resolution
f Predictors: (Constant), System Navigation, Content, Problem Resolution, On-Time Delivery
g Predictors: (Constant), System Navigation, Content, Problem Resolution, On-Time Delivery, FMS Integration
The data analysis provides initial evidence for the reliability and validity of the Perceived EPQ components. The fact that two independent measures of the construct are highly correlated, and that data enables accurate prediction of OEPQ Ratings, are both indications of the potential value of the components. As noted in the introduction to this section, the analysis based on interview data is undertaken to gain a sense of the likely value of the proposed variables. Only after Phase 3 analysis can one make more confident claims regarding the reliability and validity of these variables in measuring Perceived EPQ.
Summary
This chapter illustrates the progression by which broad ideas from the literature have been explored and refined through two phases of interviews. Phase 1 interviews have been instrumental in delineating Perceived EPQ, whilst axial coding has helped to identify a more usable set of variables for further work. Additional interviews in Phase 2 have allowed the refinement codes and produced a set of selective codes. These codes have been examined in detail and their relationship with antecedent literature examined. Finally, the proposed set of components has been tested based on descriptive data from Phase 2 interviews. This analysis indicates that the set of variables appears to be a useful measure of Perceived EPQ. Based on this analysis, a measure of Perceived EPQ is proposed – the EPQ Scale. This scale incorporates the 33 selective codes derived from Phase 1 and 2 empirical analysis. Figure 26 shows
Exploring Perceived EPQ. Alistair Brandon-Jones
how the proposed scale fits within the broader EPQ Model. The model incorporates work from the literature (See figures 5, 6, 7 and 10) with empirical analysis from Phase 1 and 2. Beyond the two broad sub-sets of ‘system’ and ‘support’, the researcher does not make assumptions regarding the underlying structure of Perceived EPQ at this point of the study. The next chapter presents empirical data analysis from Phase 3 of the study. This work focuses on examining the structure of Perceived EPQ, validating the proposed EPQ Scale, and comparing two construct operationalisations.
Figure 26. EPQ Model (Post- Phase 1 & 2 Analysis) ‘System’
• FMS Integration
• Invoice Reconciliation
• System Configurability
• Reporting Capability
• Processing Complex Orders • System Security • System Availability • Screen Loading • System Navigation • Visual Appeal • Loaded Suppliers • Loaded Catalogues • Ease of Search • Order Processing • Ease of Authorisation • Orders to Suppliers • Order Lead-Time • On-Time Delivery • Order Accuracy ‘Support’ • Support Availability • Support Reliability • Support Responsiveness • Knowledge
• Talking Users’ Language • Support Flexibility • Problem Resolution • Confidentiality • Friendliness • Concern Shown • Timely Training • Appropriate Training • Information Provision • Encouraging Feedback Perceived E-Procurement Quality EPQ
Expectations Perceptions EPQ Compliance Compliance Contract System Transaction Costs Procurement Expenditure Purchase Prices
Exploring Perceived EPQ. Alistair Brandon-Jones