Any statsiticians out there??

[raaizin]

On a BMW forum there is a poster who replaced 4 fuel pumps in his 535. BMW claims a 2% failure rate. What are the odds of 1 person having 4 failures with a 2% failure rate.

My guess it would be 2/100 X 2/100 X 2/100 X 2/100 or 1 in 6.25 million

Am I incorrect??

 

[Pablo]

Originally Posted By: raaizin
On a BMW forum there is a poster who replaced 4 fuel pumps in his 535. BMW claims a 2% failure rate. What are the odds of 1 person having 4 failures with a 2% failure rate.

My guess it would be 2/100 X 2/100 X 2/100 X 2/100 or 1 in 6.25 million

Am I incorrect??


If the failures were all one lot, which his supplier may have and hadn't purged.....this would make such calculations irrelevant.

 

[pickled]

And what is the probability that the installer or vehicle is at fault?? Sometimes the other probabilities are more interesting to calculate LOL!

 

[tom slick]

2% failure rate is terrible!

I wonder how that 2% is measured? 2 per 100 pumps? 2 per hundred vehicles delivered?

Many companies that are QA concerned aim at 6 sigma or 6 sigma minus 1.5. 6 sigma is 3.4 defects per million or .000034% defects.

 

[Pablo]

Originally Posted By: tom slick
2% failure rate is terrible!



Didn't want to point out the obvious!!

 

[CivicFan]

Originally Posted By: raaizin
On a BMW forum there is a poster who replaced 4 fuel pumps in his 535. BMW claims a 2% failure rate. What are the odds of 1 person having 4 failures with a 2% failure rate.

My guess it would be 2/100 X 2/100 X 2/100 X 2/100 or 1 in 6.25 million

Am I incorrect??


A statistician would want to see a large sample - as large as possible. One person is not a sample and cannot make a hypothesis out of it.

Such an experience would be an outlier.

 

[pickled]

Agreed that 2% failure rate sucks. Sounds like the fuel pump manufacturer needs to dust off MIL STANDARD 105-E or the actual current ANSI/ASQ z1.4 and perform a few lot inspections while they go off and implement some process improvements. As one of my colleagues once stated, "Take Pride in Your Rework Gentlemen!".

 

[adam123]

A 2% failure rate, assuming (And I have no idea how many there actually are) that 6 million exist, 120,000 failures in a given time period. As for the odds, I would think it would be, .02*.02*.02*.02 or a .00000016/100 chance of that happening, just got back from the night shift so take my numbers with a grain of salt.

 

[CivicFan]

It doesn't mention what time frame 2% failure is - during the past century, during the warranty period or what? Is it per owner, per car or per fuel pump. So many questions... Lucky for me, I don't care much to worry about the precision of such statistic.

 

[crinkles]

i would say it is pretty complex, and not a simple task, as someone pointed out, you need the time element. their 2% would be tied to a time period, and by then you are looking at combinations of probability distributions. for us folks the best we can do is say that at any point in time the pump had a 2% chance of failing.