[R] Non-normal data issues in PhD software engineering experiment

[Gustaf Rydevik]

On Thu, Jul 10, 2008 at 5:15 PM, Andrew Jackson <ajackso at cs.tcd.ie> wrote:
> Hi All,
>

Hi Andrew,

The main questions here are not R-related, but statistical modelling
questions, and much too broad for the R list. They are things you'd
ask a (paid) statistical consultant. I would suggest taking contact
with your own university's statistical support unit:

http://www.insightsc.ie/statistics_clinic.htm
,and discuss the best approach for analysis.

Rummaging around in R, looking for tests that you can squeeze your
data into *really* isn't the best approach (and ?friedman.test clearly
states that it's for unreplicated designs only).

Some things I'd  want to know if I were your statistical consultant:

-What are you doing with your research? what's your goal?

-What exactly are sensitivity
(,"coverage","execution","infection","propogation") measuring? looking
at your data, it seems as if sensitivity is making discrete jumps. Why
is this?

-What's your actual hypothesis? That the means sensitivity values for
the two paradigms differ by a constant no matter the version? or that
they differ by a fraction?

- If this is measured on test persons, I assume that you used each
person several times. Is that so?

Answers to the above questions might be good to bring to your meeting
with the statistics faculty.


Good luck with your research,

Gustaf
-- 
Gustaf Rydevik, M.Sci.
tel: +46(0)703 051 451
address:Essingetorget 40,112 66 Stockholm, SE
skype:gustaf_rydevik