maemo.org - Talk - View Single Post

	Texrat	2010-01-22 , 01:17
	Posts: 11,700 \| Thanked: 10,045 times \| Joined on Jun 2006 @ North Texas, USA	#148

Originally Posted by MontyBravo

Your a data analyst and your honestly state this as a factual basis ? shocking!

Let me explain my comments.

Non reports do not dictate the amount of product errors. Yes complaints will become more forthcoming in fault based threads. There is a user barrier to starting a post being negative in a community based on the device. One someone has started with faults others will feel more at home in discussing this.

Anyone that reads the forum will note that any negative comments are bashed (as well as being sensibly discussed by forum members).

You are of course correct that the user experience will be dictated by expectations (well actually im putting that out there to be honest lol),

BUT!

simply put again , just to make it clear,

this is not the only reporting channel and the large majority of owners will not be reporting in this forum.

Its also the case that only a small amount of people which actually have faults will bother to report them on here.

that means that a small number of people reporting faults have a greater statistical importance and weight.

So to summarise your point I have quoted is not correct. I am sure since you are a data analyst you understand "weighted" data and the impact this has. This is what I am referring to. Hopefully i am using the correct terminology? I would imagine the meaning is clear.

And I do agree the data means a lot more with tight data control, but this does not invalidate user response on this forum.

I do understand you get lots of trolls on here. I am actually not one of them , please feel free to explain how my logic is incorrect. Not in one reply to me have you attempted to do this. And I dont need you to preach but thanks for the thought

cheers,

MB

Your logic is incorrect because without the context I've already mentioned the isolated, unqualified number of raw complaints tend to be near useless. And I'd love to know where you learned what you base your position on, because again, it's completely in error. You bounce a statement like "a small number of people reporting faults have a greater statistical importance and weight" off a six sigma black belt and you're in for a lecture.

Non reports indicate the norm. Without the norm, you don't know defect rates. Best way to get the norm of course is getting at production data and distribution data. Without that, you need to conduct a scientific study that takes into account usage, demographics and other data. Nothing remotely resembling a scientific study going on here. Valid complaints maybe, but unscientific as a "study" and not robust enough for analysis.

And the smaller your sample size, the less reliable your assumptions. If 10 devices are involved in your original lot (which in this case is worse: an unknown), then yes, 5 is an indicator of something seriously wrong. If it's 100, it's something worth investigating. If it's 1000, you've probably got outliers. Continue ad nauseum.

You keep wanting to make the case that X number of complaints out of an unknown at-large quantity not only represent the norm (which is crazy enough), but you seem to be saying that the smaller the X, the better!

Not buying it. And no, weighting doesn't help your case, sorry.

References:

http://www.springerlink.com/content/r60k11l76u5816ku/

http://www.isixsigma.com/forum/ask_d...stId=5&catId=6

http://www.sixsigmaspc.com/dictionar...litylevel.html

http://www.aafes.com/qa/docs/AQL_Tab...l_2.5_norm.htm

http://en.wikipedia.org/wiki/Failure_rate

__________________
Nokia Developer Champion
Different <> Wrong | Listen - Judgment = Progress | People + Trust = Success
My personal site: http://texrat.net

Quote & Reply |