Statistics 200 Honors
Elementary Statistics
Fall 1999

Thursday, August 26

Thursday, September 2

These data were collected from 227 respondents to a survey (out of 240 asked to fill out the survey) in a past section of Stat 200.

By the end of this computer lab, you should be able to do the following using MINITAB:

1. Simulate a binomial variable (and use this to simulate a sample proportion).
2. Obtain entries in the standard normal table at the front of your book.
3. Produce a histogram or boxplot of a univariate distribution.
4. Recode a categorical variable from numeric to text.
5. Obtain a confidence interval for a proportion.
6. Produce a cross-tabulation table for two categorical variables.

• Quantitative variables: Numeric variables for which arithmetic operations such as adding and averaging make sense.
• Categorical variables: Variables which place an individual into one of several groups or categories. These may be numeric, but the numbers only signify which group an individual belongs to.

Thursday, September 9

I:\STAT\200\Old_Files\Survey99.MTW

I have written one solution to the homework question, which you can see in pdf or postscript format. Please don't use the same two variables I used!

Here are some dangers of this assignment:

• Make sure you choose two categorical variables. Quantitative variables will not work.
• Try to choose variables without too many categories. You don't want the table to get too sparse, meaning that you don't want small numbers in any of the cells if you can avoid it. Thus, by the same token, don't choose variables in which one of the choices is likely to have very few people who choose it.
• As a rule of thumb, none of your expected (not observed) counts should be less than 5 in order that the chi-square test is approximately a correct test to use. The test is more accurate for larger samples, and 5 is just a generally accepted cutoff for the lowest expected count. If you choose variables which result in counts lower than 5, you might want to choose new ones or talk to me about it.
• Finally, don't forget that you should explain any Minitab output you turn in--but I encourage you to turn in such output.

Thursday, September 16

I:\STAT\200\Old_Files\Survey99.MTW

I have written one solution to the homework question, which you can see in pdf or postscript format. Please don't use the sports participation variable that I used.

The assignment requires that you use a test comparing two proportions and a cross-tabulation chi-square test. In addition, you may want to recode the gender question and your yes-no question of choice from numeric to text. Here's how to do each of these things:

1. To recode (for example) variable C1 so that 0 becomes Male and 1 becomes Female, select Code from the Manip menu. Choose numeric to text. Both "from columns" and "to columns" should be C1 unless you want to create a new column C228 and leave the original C1 intact. Enter the original values (0, 1) and the new values (Male, Female). If you prefer to type the commands, see the solution to find out what commands to type.
2. To run a two-proportion test, choose Basic Statistics from the Stat menu, then choose 2 proportions. You want "samples in one column"; the samples variable is the one you've decided to study and the subscripts variable should be C1 (gender). Before you run the test, make sure you click Options and check the pooled estimate option. If you don't do this, the test you perform will be slightly different from the one specified on p. 605.
3. To run a cross-tab test, choose Tables from the Stat menu, then choose Cross tabulation. Be sure to check the chi-square box. Also check the row and/or column percentage boxes. This will produce not only the table you want, but the chi-square statistic and its p-value as well.

Thursday, September 23

Thursday, October 14

• Know what a scatterplot is and how to create it on Minitab.
• Know what the correlation r measures and how to compute it on Minitab.
• Do exercise 2.19, p. 131.

Thursday, October 21

Here is a list of the Minitab tasks at which you should try to be somewhat proficient before you leave today:

• Plotting scatterplots, with or without regression lines added.
• Computing the equation for a regression line, the correlation between two variables, and the value of r-squared.
• Using the equation to predict values of the response variable for a given value of the predictor variable.
• Obtaining a column of residuals and making residual plots of various types.
• (*) Obtaining confidence intervals for predicted values, the slope of the regression line, and the intercept of the regression line.
• (*) Obtaining sum of squares, degrees of freedom, mean squares, and an F statistic for the linear regression model.

The help menu may be of interest to you as you attempt these tasks. Be sure to explore each of the following menu choices at some point today:

• Graph - Plot
• Stat - Regression - Regression
• Stat - Regression - Fitted Line Plot
• Stat - Regression - Residual Plots

Thursday, November 4

The significance tests for individual regression coefficients assess the significance of each predictor variable assuming that all other predictors are included in the regression equation.

In order to do this, you'll formulate a simple model which includes two predictors. Each predictor is strongly linearly correlated with the response variable. However, neither of them is significant when you look at a regression model with both thrown in.

Can you see how to do this?