<p>I'm pretty sure there's going to be a lot of 2-sample tests.</p>
<p>When doing a chi-square test, do you have to write what it's for (ie independence, homegeity, etc.?</p>
<p>I thought that uou'd be looking for "goodness of fit" ie. does the expected proportion coincide with the observed proportion when using Chi Squared?</p>
<p>Yes you have to identify it as Goodness of fit, independence, or homogeneity</p>
<p>Just say why you would use it</p>
<p>So if you don't, do they assume you didn't name the test and take off a fraction of a point?</p>
<p>So what are all the possible occasions where you use the chi-square test?</p>
<p>You should be clear as to why you are using it, and the assumptions that goes along with the test. So I guess that they will deduct marks if you do not fill out that requirement.</p>
<p>Is there going to be questions on transformations?</p>
<p>They grade you on a 1-4 scale, so if you're partially correct on a part or omit an insignificant thing, you can still get a 4, depending on the error and the question itself.</p>
<p>Ok so let me see if I got this right .</p>
<p>Both type I and II errors decrease as n increases and confidence intervals decrease.</p>
<p>Type II increases when type I decreases and vice versa.</p>
<p>Power of test increases as n increases and also when type I increases.</p>
<p>Is this correct?</p>
<p>Also, what other purposes does a (alpha) serve?</p>
<p>Err, Type I error probability = the alpha level. It only changes with the percentage size of the confidence interval.</p>
<p>I thought that if you had a greater n, the Type I error is also less :confused:</p>
<p>No, Type I means rejecting a true mean. Think about when you reject a mean with a confidence interval; it's when you get results outside the confidence interval.</p>
<p>Chi Square Test</p>
<p>Goodness of Fit: An observed pattern of data fits come given distribution
Independence: Two or more samples came from a larger set, meaning they are independent of each other</p>
<p>The chi-square test of goodness of fit is used to test the hypothesis that the total sample N is distributed evenly among all levels of the relevant factor.</p>
<p>The chi-square test of independence is used to test the null hypothesis that the frequency within cells is what would be expected, given these marginal Ns.</p>
<p>To do the goodness of fit test, they will usually give you a list of observed and expected values, and you have to test whether or not there is a good fit. (Enter observed in L1 expected in L2, L3 = (L1-L2)^2/L2 </p>
<p>To do a test of independence, they will usually give you a chart with two or more categories. You have to calculate the expected values (you can do that by entering the data values into a matrix, then running the X^2 test to get your expected values) </p>
<p>Lower x2 value -> Better fit</p>
<p>when comparing two samples by t, or z test, do you always take the difference between the two means?
I'm confused because it seems that while the means are being subtracted from each other, the standard deviations in contrast, are being added together under the square root? Is this true? Or are my notes wrong?</p>
<p>You always add the variances, considering variance measures spread.</p>
<p>Thanks Ray. </p>
<p>So what exactly are matched pairs?</p>
<p><a href="mailto:libri-crucis@satx.rr.com">libri-crucis@satx.rr.com</a></p>
<p>any past Stats exams to give me please? :)</p>
<p>Matched pairs is one subject, two treatments.</p>