Is Pearson's chi squared test the right method? I have a sample of n=1000. The sample covers cars being brought in for service after one year of ownership in my country.  For each car, I know which defects it had when it was brought in. I'm trying to work out whether the occurrence of errors, which are specific for my country, are deviating from the global numbers or not.
My statistic covers 10 different defects, meaning that I have the actual percentage of cars which had a certain defect and the global percentage of cars which had the same defect. 
Hence, I'm essentially comparing two columns, 10 numbers each.
Doing a CHISQ.TEST in Excel doesn't seem to take the sample size into account. Am I using the wrong test here or is the test valid for this purpose?
 A: You have ten categories, and so nine degrees of freedom (DF) for
the approximately chi-squared test statistic. It is true that DF
does not depend on the sample size. However, the sample size
is not irrelevant to the power of the test. 
I suppose you are taking the 'global' proportions as the hypothetical
model and using the chi-squared goodness-of-fit test to check
whether data for your country match the hypothetical model.
This is a little like a test to see if a die is fair. This is a
 reasonable procedure, and I would be happy to check your results. (In the case of a die the hypothetical model would ordinarily
have equal probabilities for each of six categories, and in
your application, categories would not necessarily have equal probabilities.)
However, in the dice example, n=60 rolls would not be enough reliably
to distinguish a biased die (specifically where faces 1, 2, and 3 each have
probability 1/12 and faces 4, 5, 6 each have probability 3/12) from
a fair one. By contrast n=600 rolls would give much better power
(probability of rejecting null hypothesis of fairness when die
is biased in this manner). This influence of sample size on properties of a chi-squared test can be shown by simulation. (I will do it, if you are interested.)
