Assignment for Monday: Create an outline of the key points in the chapter, including all vocabulary words.
Also, do problems 4.54, 4.56, 4.62 (this study just celebrated its 20th anniversary!!!), and 4.66
So, how about those marginal and conditional distributions for two-way tables, huh?
The marginal distributions are the percents that each column or row represents in the entire table. For instance, if the total of one row was 250 and the total for the table was 1000, the marginal distribution for that row is 25%. You would continue to calculate percents for all of the other rows or the other columns--whatever the question asked for.
For the conditional distributions, you only consider a portion of your population, for instance only a specific row or column. Then, what portion of the observations recorded in tht small group shared the desired characteristic?
If there were 15 sophomores taking AP Chinese and 500 sophomores in a school of 2000 students, GIVEN THAT a student is a sophomore, the percent who are taking AP Chinese is 100*15/500.
Tuesday. October 16
Assignment for Friday: 4.34, .36, .38, .39, .40, .42, and .43
How did you like the Simpson's Paradox activity today?
When breaking data into two or more divisions by a lurking variable changes the "decision" for EVERY ONE of the sub-groups, the result is a Simpson's Paradox. For instance, the example today presented no clear, justifiable answer about whether we should fund Bolgg's Panacea or not.
The example in the book about the hospitals is instructive.
Good luck on the PSAT.
Monday, October 15
4.22-4.24, 4.27, and 4.28
Review the topics and procedures on the notes handed out today.
Have you ever heard of Simpson's Paradox???
Thursday, October 11
We've transformed non-linear data to a linear form, found the LSRL through the data, re-written the equation reflecting the nature of the lists used to develop the LSRL, and re-transformed the equation to model the original data.
You should have done problems 4.6 and 4.9. For tonight, DO problem 13 and READ ACTIVITY 4 and problem 4.15. If you feel excited about the investigation, read problem 16 also.
Monday, October 8
We graphed some relationships between x and y to determine whether we were allowed to run the LSRL on the data. Of course, we ONLY run the least squares regression on data that look like they have a linear pattern.
When the pattern in L1 and L2 looked like an exponential growth or decay model, we took the log of y in order to un-do the exponential. Putting the log y into L3, we proceeded to verify that the graph of L1 and L3 was approximately linear. We then ran the LSRL through that set of points.
The equation we found by using the LSRL will not run through our curve-y data, so we have to un-transform the equation. For the exponential case, we had used the log of y instead of y itself when finding the LSRL (but the original x values!), so we re-write the equation as log y-hat = a + bx.
We solve for y by taking the antilog of both sides ("ten-to-the" or 10^stuff). The resulting equation for y can be graphed with the original x and y data and should match the pattern pretty well.
If the model looks like a quadratic, square root, or other power function, you'll need to perform mostly the same functions, but on the logs of both x and y. The linear equation that passes through the straightened data will be transformed like this: log y-hat = a + b times log x, and the result will have a factor equal to x to the b power.
HW: Problem 4.1. The answer is in the back of the book, but there are a lot of sections to this problem.
Posts
