|
|
I had posted a question last week about analyzing categorical/ordinal
data, but have a slightly different question this time.
I am looking for relatively simple ways to display pre and post test
differences in a survey that uses a likert type scale. We have done a
reliability analysis and the items have high reliability. We have
also done a factor analysis and came up with 12 factors (based on 100
+ questions).
I want to show how responses to groups of items differed pre and post
(not trying to make claims about causality, but do need present the
information to the course instructors).
Is it problematic to aggregate responses to the related items to
simplify reporting on each factor/group of items? This is what I had
in mind....
FACTOR 1 PRETEST
NONE LOW MOD HIGH VERYHIGH TOTAL N
Item 1 2 5 23 4 2 36
Item 2 4 5 21 5 3 38
Item 3 etc......
TOTAL N 6 10 44 9 3 72
TOTAL % 8% 14% 61% 13% 4%
FACTOR 1 POSTTEST
NONE LOW MOD HIGH VERYHIGH TOTAL N
Item 1 0 2 20 10 4 36
Item 2 2 3 15 9 9 38
Item 3 etc......
TOTAL N 2 5 35 19 13 72
TOTAL % 3% 7% 49% 26% 18%
I could then create pie charts for pre and post and one could see how
the response breakdowns differ pre and post....
If there are problems with this way of analyzing and presenting the
information, could you explain what the drawbacks are?
Thanks,
Angela
|
Locked
