Likert scale - questionnaire administered twice - longitudinal study

#1
Hi,

I am running a longitudinal study which lasted for few months during which
all participants used a specific communicaiton device. I conducted a questionnaire
at the beginning and the same questionnaire was administrated at the end of
the study. All participants performed the same protocol (they
are all using the device). The questionnaire was administrated twice to see if
their answers is going to change because of the device.
The questions are all a 5-level Likrt Scale.


In the analysis; I would like check two things:
1) if there is statistical difference between the answers pre/post study
2) if age and frequency of using the new device has an impact

My data is organized as:
Participant Age Usage Q1Pre Q1Post.......... SumPre SumPost Delta
1 23 45 3 4 20 23 3
...
....

where Delta is SumPost-SumPre


For (1) I know I need to use the Wilcoxson signed rank test since the data is
ordinal (please correct me if I am wrong).

For (2), I am confused! My first question is ; can I treat
SumPre, SumPost and Delta as quantitative variables even though they are extracted
from ordinal/likert scale?

Second question: I could do it one of several ways but not sure which is right (if any:)

(A) Use Age, Usage and SumPre SumPost
variables: the above variable looks (to me) quantitative, quantitative, quantitative, quantitative
analysis: which I guess means I should be going for Spearman Correlation linear Regression


(B) Use Age, Usage, Delta
variables: The above variable looks (to me) quantitative, quantitative, quantitative
analysis: same as above (I guess)

Or I could introduce categorical variables: AgeGroup and UsageGroup

(C) Use AgeGroup, UsageGroup, SumPre and SumPost
variables: The above variable looks (to me) categorical, categorical, quantitative, quantitative
analysis: not sure?


(D) Use AgeGroup, UsageGroup, Delta
variables: The above variable looks (to me) categorical, categorical, quantitative
analysis: not sure?


I'd appreciate your advise which of A/B/C/D to use and which technique
to use with it?

FYI, I am expecting to see that age and frequency of usage has an effect on the questionnaire result that was administrated at the end of the study.

Best.
Anya
 
#2
Just a continuation to my post:

I am inclined now to use ordinal regression with Age,Usage as covariates and SumPost as dependent variable.
But SumPost is the sum of each score from each Likert question and I am treating it this way as scale though it is really
ordinal (or is it scale?) Can I actually sum the responses from a likert questionnaire so that every participant has a single score?? Statistics confuses me :)

Hope someone answers!