Hi all,
The 95% figure comes out of thin air - if you'd like to adjust it to
99%, and recalculate the overall odds of correctness, that would be
fine. In fact, I believe the original version of the text was something
like:
"...if there's a 95% chance of each line being correct, the odds of the
whole function being right are only 41%. Even if 99% of lines are
correct, the odds are only 84%, which is still uncomfortably low."
Do you think that would be clearer?
Cheers,
Greg
On 2014-11-24 1:10 PM, Shoaib Sufi wrote:
I didn't assert that it was the point.
Thanks for the reference.
On 24 Nov 2014 17:51, "Aron Ahmadia" <[email protected]
<mailto:[email protected]>> wrote:
From McConnell's summary in Code Complete:
> Industry average experience is about 1-25 errors per 1000 lines
of code for delivered software.
Of course, that's production code that has presumably been through
far more testing and other forms of quality assurance than
scientific code. The 95% number is just an assumption for an
example, not a precise citation. Greg may know of a better number
for scientific code, but I don't think that's really the point of
the exercise.
On Mon, Nov 24, 2014 at 12:44 PM, Shoaib Sufi
<[email protected]
<mailto:[email protected]>> wrote:
Hi Aron,
What I mean is where does the assumption of a line of code
being 95%
correct come from.
Thank you for helping me think more clearly about the question
I wanted to ask.
Best
Shoaib
On Mon, Nov 24, 2014 at 5:38 PM, Aron Ahmadia
<[email protected] <mailto:[email protected]>> wrote:
> It's a function of statistics, assuming each of the lines of
code is an
> independent distribution that is either correct or wrong.
>
> Given the input assumption (95% of all source code lines are
correct as
> written the first time), then the code is correct if the
individual lines
> are all correct, which has probability P = 0.95^17. This is
more correctly
> rounded to 42%, but it's in the right ballpark :)
>
> On Mon, Nov 24, 2014 at 12:34 PM, Shoaib Sufi
<[email protected]
<mailto:[email protected]>>
> wrote:
>>
>> Hi,
>>
>> At the bottom of:
>>
>>
https://github.com/swcarpentry/bc/blob/gh-pages/novice/r/04-cond-colors-R.Rmd
>>
>> It states:
>>
>> 'Our final heatmap function is 17 lines long, which means
that if
>> there's a 95% chance of each line being correct, the odds
of the whole
>> function being right are only 41%. Before we go any
further, we need
>> to learn how to test whether our code is doing what we want
it to do,
>> and that will be the subject of the next lesson.'
>>
>> Where is the reference for making a statement like that -
i.e. %
>> chance of errors based on function length.
>>
>> Thanks
>> Shoaib Sufi
>>
>> _______________________________________________
>> Discuss mailing list
>> [email protected]
<mailto:[email protected]>
>>
>>
http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
>
>
_______________________________________________
Discuss mailing list
[email protected]
http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
--
Dr. Greg Wilson | [email protected]
Software Carpentry | http://software-carpentry.org
_______________________________________________
Discuss mailing list
[email protected]
http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
