Importing a BMW from the US is nonsensical - it will be more expensive than buying one locally!
Steer clear of second-hand BMWs
- Thread starter Sefirosu
- Start date
![[TW]Fox](/data/man-of-honor/2009.png){kind=avatar}
No, because the Americans didnt have 328i's in 2003 either.
Soldato
Over here.
[TW]Fox;22394070 said:
I know that, but don't you have to at least have a suitably sized representative sample for the results to mean anything?
I take ten men I ask them if they like cheese 6 of them say no, therefore 60% of all men on the entire planet don't like cheese?
40,000 car owners means you could have just asked 40000 @@@@witts who fill their engine with sand entirely by accident.
I probably know about, I don't know say I know 50 people relatively well.
None of them own BMWs.
Therefore, NOBODY knows BMWs?
Last edited:
40k is a good sample unlike Pantene 53/126 agree
PROVIDING YOUR SAMPLING IS APPROPRIATE
'People atpbx knows' is not an appropriate sample pool for ascertaining average BMW ownership levels.
Soldato
Does this mean I should buy a Thrust SSC cos they've had the least breakdowns?
Caporegime
*fewest
Associate
You take a sample representing 0.15% (in this case the 40,000 they asked in the UK) and equate it to your "cheese" survey representing 0.00000014%, where one survey's sample size uses over 1 million times the population it's testing? Not to mention the fact the 50 people you know are far more biased than the 40,000 they used in the survey - simply because they know you. Most of the 50 people you know probably come from the same area as you, work in similar places (I can probably assume quite a large chunk of these 50 people work with you) meaning they're probably mostly earning around what you earn and countless other similarities.
Its taking it to extremes i know.
But 40k sample size of a 26.5 million unit user base doesnt tell you anything about anything.
Associate
You realise they didn't just pluck that number out of the sky? Calculations involving confidence levels, error rates would've been performed to establish a sample size not worth going beyond. A sample size of 0.15%, while small, can still give very accurate representations of the overall population.
40,000 is probably fine. You could survey 5,000,000,000 people accurately with a 40,000 sample if you chose the sample correctly for the subject at hand. In the UK the regional, class, etc. variations are probably not significant enough to have an impact on a cars 'Mean Time To Repair' that it warrants a sample any larger. 40,000 allows you coverage of most models in most areas with most types of owner.
Where I think there is probably a weakness is in the choice of car - if you want a '5 Series', do you pick a 520i or an M5? Both have very different potential repairs and thus repair times. I'd like to know what methodology was chosen. Highest sales volume of the type per model would be sensible from the perspective of usefulnes, but would paint an unfair picture to some degree.
The crux of this tends to be, when things like BMWs top the list, is that 'more complex cars take longer to repair'. I don't think anyone is surprised by that.
Where I think there is probably a weakness is in the choice of car - if you want a '5 Series', do you pick a 520i or an M5? Both have very different potential repairs and thus repair times. I'd like to know what methodology was chosen. Highest sales volume of the type per model would be sensible from the perspective of usefulnes, but would paint an unfair picture to some degree.
The crux of this tends to be, when things like BMWs top the list, is that 'more complex cars take longer to repair'. I don't think anyone is surprised by that.
Last edited: