17.01.2013 Views

Cash or Card: Consumer Perceptions of Payment Modes - Scholarly ...

Cash or Card: Consumer Perceptions of Payment Modes - Scholarly ...

Cash or Card: Consumer Perceptions of Payment Modes - Scholarly ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

6.3.3: Pre Test One- Internal Consistency Assessment<br />

Reliability f<strong>or</strong> internal consistency is based on c<strong>or</strong>relations. The internal consistency <strong>of</strong> the<br />

scale is evaluated via examining inter-item c<strong>or</strong>relations, items-to-total and Cronbach alpha.<br />

Acc<strong>or</strong>ding to De Vellis (2003) internal consistency reliability is the measure <strong>of</strong> homogeneity<br />

<strong>of</strong> the items within the scale. The logic is that the relationships among items are logically<br />

connected to the relationships <strong>of</strong> the items to the latent variable; theref<strong>or</strong>e, high inter-item<br />

c<strong>or</strong>relations suggest that items are all measuring the same thing. De Vellis (2003) and<br />

Netemeyer, et al. (2002) suggest evaluating the item-to-total c<strong>or</strong>relations f<strong>or</strong> internal<br />

consistency assessment. An item-total c<strong>or</strong>relation test checks whether an item is consistent<br />

with the average behaviour <strong>of</strong> the other items in the scale. A Pearson’s c<strong>or</strong>relation below .30<br />

indicates low c<strong>or</strong>relation. The most common test f<strong>or</strong> internal consistency is Cronbach’s<br />

Alpha (Netemeyer, et al., 2003). C<strong>or</strong>tina (1993) cautioned not to apply alpha when assessing<br />

dimensionality because interrelatedness <strong>of</strong> items within a scale does not imply<br />

unidimensionality (the existence <strong>of</strong> one latent construct underlying a set <strong>of</strong> items) (Hattie,<br />

1985). What constitutes minimum acceptable alpha level is debated, however, “a widely<br />

advocated level <strong>of</strong> adequacy f<strong>or</strong> alpha is.70” (Netemeyer et al. 2003, p.58). EFA and scale<br />

reliability analysis shows that not all dimensions were equal <strong>or</strong> above 0.70 levels alpha but<br />

overall average alpha value meets the acceptance (See Table 6.2).<br />

The c<strong>or</strong>relations within dimensions (positive emotions, negative emotions, control spending,<br />

status/pleasure, money as gift and behaviour/usage) show all items are c<strong>or</strong>related moderately<br />

(See Table 6.5). There is no clear guideline to what constitutes moderate inter-item<br />

c<strong>or</strong>relations. F<strong>or</strong> example, Robinson, Shaver, and Wrightsman (1991) advocate average inter-<br />

item c<strong>or</strong>relations <strong>of</strong> .30 and Peterson (1994) ask f<strong>or</strong> .31 and an overall average alpha <strong>of</strong> .80.<br />

Clark and Watson (1995) consider average inter-item c<strong>or</strong>relations <strong>of</strong> .15 to .50 across<br />

construct to be acceptable. The items-to-total c<strong>or</strong>relation (the c<strong>or</strong>relation <strong>of</strong> the item and the<br />

sum sc<strong>or</strong>e <strong>of</strong> the other items) show all items within dimensions are internally consistent.<br />

C<strong>or</strong>relations <strong>of</strong> less than 0.30 f<strong>or</strong> item-to-total c<strong>or</strong>relation are considered weak (de Vaus,<br />

2004). The “behaviour/usage” fact<strong>or</strong> shows weak c<strong>or</strong>relations between the items, less than<br />

.20 and below. These items were retained at this stage because they identified payment mode<br />

perceptions.<br />

95

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!