Must all equal values be in the same quartile

[alexbib]

Posted this in the stats section, but nobody seems to be reading that section...

I have a few questions about distributing values into quartiles:

1.must all equal values be in the same quartile, even if that makes you create (sometimes very)unequal groups?

2.what do you do when your number of values cannot be divided by 4? Say I have 10 values, do I make the quartiles 3-2-3-2, 2-3-3-2, or something else?

About percentiles, what do you do when you have less than 100 values?


I know stats isn't meant to be used on small numbers of values, but I am tutoring two high school students and couldn't answer these questions.

Thanks,

Alex

 

[mathman]

I believe you are trying to make a mountain out of a mohill. Specifically, to have EXACT quartiles, the number of samples has to divisible by 4, similiarly for percentiles, divisible by 100. Otherwise you are stuck with approximations.

 

[selfAdjoint]

In addition to what mathman said, you shouldn't try to make your quartile "look like" some preconceived idea. For small samples, you can't expect to see any normal curve behavior. And perhaps your data doesn't fall on a normal curve anyway. Best to use software to do your binning to avoid operator bias.

 

[alexbib]

Unfortunately, I cannot use software. I have these questions because I am tutoring high school students who have written tests about dividing small samples into quartiles.