On Sun, 11 Oct 2020 13:45:47 -0400, Rich Ulrich
Post by Rich Ulrich
On Sun, 11 Oct 2020 06:23:15 -0700 (PDT), Eshani Sandali
Post by Eshani Sandali
When computing the variables, should i keep the
values for the computed variables as same as values
that have been added for the individual questions?
(EX= 1= strongly disagree..) and what should be
the measure of computed variable?
Sorry. I can't make sense of the question(s). Computing
what variables? And what is meant by "values ... added"?
If you want to get the Predicted value for each case in
a multiple regression, request it as an option.
** Here is the response from Eshani, mistakenly sent to my
email address **
<< I have selected five independent variables namely, perceived
usefulness, ease of use, risk, trust and awareness. For each of the
dimensions i created five questions. When doing the multiple
regression i have to compute the five questions so that i can get a
one computed variable called Perceived usefulness and in the same
way for other variables as well. In that computed variable, the mean
value can be a in between value. So before running the multiple
regression do i have to round off them or keep it the way as the
results produced? >>
Creating "composite" or "factor" scores for each of the sets of
items is a great idea. Some people have used Total scores; I
prefer using the Average scores for test items, because using
the Means keep visible the verbal anchors for the scores.
There is absolutely no reason to truncate or round off the
item averages before using MR.
The one circumstance where I routinely round off variables
is: after computing "T-scores" -- I create those as composite
scores averaged from multiple factors, and standardized to have
a mean of 50 and SD of 10. I round them off (a) because
there is going to be essentially no loss of precision and (b)
so that I can list scores without decimals, or I can point to
individual scores and not worry about two scores that "look
the same" being different by a trivial fraction.