<p>I'm assumin u mean 2 proportions CI/sig test...</p>
<p>CI:
SRS, pop is 10X n, both n's are big so that np>=10 or 5 or so, same with n(1-p).</p>
<p>For sig test:
same thing, except p-hat must also = ((x1+x2)/(n1+n2))</p>
<p>I got a question:
Can someone explain when to use a normal, binomial, and geometric cdf/pdf...and how to enter it into calculator???</p>
<p>95% confidence interval means that 95 out of 100 values will be contained in (insert range here)....and that we are 95% confident that the true mean/proportion of (insert problem context here) lies between (insert rang)</p>
<p>to r1400sch</p>
<p>normal is used only when you have z scores it is normcdf(z,99) <you never really need to use pdf</p>
<p>geometric is when you want the probability of only ONE sucess. Geopdf is for on specifically the nth trial geopdf(P sucess, nth trial) if you want a sucess WITHIN n trials (ie n or less) then use cdf Geocdf(P sucess, nth trial)</p>
<p>Binomial is P of x sucesses within n trials again pdf gives exactly and cdf is within. key order is binomialpdf/cdf(n,P,x)</p>
<p>just a quick ?
when a prob asks you to find an sample size with 95% confidence so that the margin of error is 5...
m = z* x sigma/square root of n...</p>
<p>does m = 5 or 2.5 (2.5 on each side makes the total 5)?</p>
<p>CI = test statistic +/- Z* times standard deviaton or SEM
the part after the "+/-" will equal 2.5</p>
<p>as far as CI's
I suggest not using the phrase "lies between"
instead, use captures or contains</p>
<p>only because the former suggest that the true parameter is moving within the range when actually it is the rang that is moving to "capture" the true parameter</p>
<p>To Dpat, you usually don't say margin of error is 5 when its actually +/- 2.5, you say that the margin of error is 2.5.</p>
<p>wait .. but if the problem asks him to find the sample size WHEN the margin of error is 5, isn't margin of error just ... 5? He's trying to find sample size n, not margin of error. Or did I misunderstand it?</p>
<p>Lucidity, whenever they ask you to find n for a certain margin of error, You use that number as m in the equation just like you said. Dpat seemed to interpret margin of error in a different way, and that's why explained it like that in post#27.</p>
<p>In that situation you will set 5=z*std. deviation, since that is your margin of error. You will just solve for n from this equation.</p>
<p>I'm a little confused on two thigns. Which Chi test is the one with the matrixes, what's the other one (where you do List 1, list 2, than (O-E)^2/E)?</p>
<p>Also, what are the assupmtions for each?</p>
<p>Chi-square</p>
<p>You always do sum of (Observed-Expected)^/E. Observed for a table is (row toatl)(column total)/total.</p>
<p>Assumptions is that all things are bigger than 1, only 20% less than 5.</p>
<p>And if I'm wrong, someone should correct me.</p>
<p>chi-squared test for independence and chi-squared test for goodness of fit are both done using matrices</p>
<p>Not even sure how to do a test for homogeneity, i think it was #6 for one AP FR</p>
<p>Btw, I see a lot of variations between conditions for CHI-SQUARED.
Supposedly our teacher told us to use the following for all three tests:
1. SRS
2. all expected counts > 0
3. no more than 20% of expected counts < 5</p>
<p>someone verify? i'm pretty sure it doesn't apply to homogeneity but i really need to find a fully worked example of that test anyways.</p>
<p>shravas, the expected count is (row total)(column total)/(grand total) NOT the observed</p>
<p>Goodness of fit is done with two lists. Homogeneity and independence are tested using matrices (they are essentially the same test, just different explanations).</p>
<p>One assumption that I've seen is independently collected samples, not sure if that one is common though.</p>
<p>EDIT: And of course the data should be counted...</p>
<p>Snipez90--
Those are the same conditions my teacher told us to follow, but I'm looking around various Stat texts for last-minute verification.</p>
<p>This is kinda stupid, but how do you know which hypothesis should be set as the null and which should be set as the alternative?</p>
<p>Your null hypothesis is typically mu=0 or some variant of that (like mu1-mu2=0 for 2-sample tests). Your alternate hypothesis on the other hand usually has doesn't-equal, greater-than (>), etc.</p>
<p>oh god i didn't even realize there were 3 different chi tests...i am so screwed</p>
<p>eep. Do we have to know advanced linear regression?</p>