<p>So we never went over assumptions in class - I only discovered the necessity of writing them down in the midst of a last-minute cram, haha - and I've got some questions. For proportion tests and confidence intervals, I've seen two different requirements: either that np and nq have to be greater than 10, or that they have to be greater than 5 - anyone want to clarify? Also, why do the proportions have to be less than 10% of the population?
Thanks :)</p>
<p>use np and nq.
10% rule is used to assume independence if problem doesn't state</p>
<p>Yeah, I know np and nq - but do both of them have to be more than five or more than ten?</p>
<p>For AP STATS they require np > 10 and nq > 10
I've only needed to use np > 5 and nq > 5 for the Statistics chapter in Mathematics 12 (in Canada)</p>
<p>Okay, gotcha. Thanks to both of you :)</p>
<p>You know how when you do a 2-samp z test for proportions you have to form a combined proportion?</p>
<p>Do you multiply the combined proportion by the combined sample size, or do you multipy it by n1 and n2?</p>
<p>When checking assumptions for 2 - samp. proportional tests...
n1p1 > 10
n1q1 > 10
n2p2 > 10
n2q2 > 10</p>
<p>Not sure by what u mean by combining?</p>
<p>in the kaplan review book (im self-studying), it says that, for sig. tests but not confidence intervals, you have to construct a combined proportion given by</p>
<p>(X1 + X2)/(n1 + n2)</p>
<p>because of the null assumtion that p1 = p2</p>
<p>Some stats say they must be greater than 10, but other ones say that is too limiting. They say smaller samples can be significant enough. Hence the 5.</p>
<p>The 10% rule is not a real rule, it deals with the finite population correction. Meaning that the sample was selected w/o replacement from the population. The actual standard error should be s/sqrt(n) * sqrt(1-n/N)
but when n/N < .1, the sqrt is like .96/.97 so its roughly the same. You should use this when the population is not infinite, but its not really in the curriculum.</p>