forum.smartpls.com

Hi Everyone,

Are there any differences in terms of outer loadings (measurement model) across 5,7 or 10 likert scales?. Outer/items loading is the first test for accessing construct reliability (e.g. Convergent validity) to ensure that each measurement item correlate strongly with its assumed theoretical constructs (Gefen et.al., 2005). Other measurements convergent validity is ICR (>0.70), AVE(0.5),and t-statistics (>1.96).

I found a paper http://business.nmsu.edu/~mhyman/M610_A ... R_2008.pdf, but this is from a survey via telephone, and validity and reliability are out from their study.

In fact, more scale point provides more options for the respondent and could result 'negative kurtosis' in terms of spread of the data. However, since the PLS was a non-parametric analysis compare to other method, the Normality assumptions are not concerned (Chin,2010).

Any idea?

Hi Brisbane,

The answer is no, there are no differences in terms of outer loadings.
As correlations, they are standardized ( – 1 to +1)

Remembering that the first step in the PLS algorithm is the standardization [(xi – xbar)/stdev] of all indicators [mean = 0; stdev = 1], then it doesn’t matter if we have 5, 7 , or 10 points.

Best regards,

Bido

Dear Prof. Bido,

Thanks.

I try to convert from 5 scales to 7 scales , using the formula/previous work done in that paper. Just for an experiment.

Thanks again