Vol: 2, Issue: 4 2012

Page: 16-21

Study on Investor Preference for Financial Instruments

Monica Sehgal

CITC indicates whether the variable actually belongs to the construct or not. Variables showing scores lower than 0.5 are deleted, unless there is a compelling reason to keep them in the construct. Some items with CITC values over 0.5 can also be removed if the overall reliability of the construct in question improves as a result of the deletion (obtained by checking the ‗alpha if deleted scores). Reliability of constructs refers to the accuracy with which the constructs repeatedly measure the same phenomenon without much variation. The reliability of each construct in question is examined using Cronbach‘s alpha (Cronbach, 1951). An alpha score larger than 0.7 is generally acceptable as sufficient accuracy for a construct (Nunnally, 1978). After purifying the constructs one by one, we arrive at the purified scales for the constructs, each of which is sufficiently reliable. Unidimensionality is a common trait exhibited by all the indicator variables of any given construct (McDonald, 1981; and Hattie, 1985). Unidimensionality is best measured by CFA. A combination of CFA and path analysis is structural equation modeling. This is the best method of measuring the unidimensionality of any construct.