software-store-developers team mailing list archive

[Aaron Peachey <alpeachey@xxxxxxxxx>]

Hi,

Ignore my previous email since I've done some more work on this to get
a more useful starting point for us.

My new pair of scripts (http://db.tt/szCPDog) now implements the
ci_lower_bound in both ruby (verbatim as supplied) and python
(re-write of ruby version) to print the output of 3 test cases fed
into ci_lower_bound.
(This uses 0.1 as the power, which should correspond to a 90%
confidence interval - not sure why the Wilson score link says 0.1=95%
but I'm happy to be corrected as it's a long time since I've done
statistics!)

In both the python and ruby script, the output is the lower bound
Wilson score for a case of:
2 ratings out of 2 (i.e. 2x5 from 2 ratings in spec test case 1),
1 rating out of 3 (1x2 from 3 ratings in spec test case 2), and
2 ratings out of 3 (2x5 from 3 ratings in spec test case 2)

Then in python, I've written a calc_erl function which takes a list of
5 integers corresponding to the number of ratings at each level of
[1,2,3,4,5] to produce the ERL as per the spec.

So the python script also prints out the resulting ERL for the two
test cases in the spec.

To run, download the archive from the link, extract and run:
ruby ci_lower_bound.rb
python calc_erl.py

Thoughts? Questions? Issues? Next steps?

cheers
Aaron