STAT 501 – Homework 12 (Lessons 14 & 15) – Spring 2020 – Due: Apr 29
Instructions: Use Word to type your answers within this document. Then, save and upload your document to Canvas by the due date. The point distribution is located next to each question. Unauthorized distribution and/or uploading of this document is strictly prohibited.
- (5×5 = 25 points) Use the “CO2Concentration” dataset, which contains the average carbon dioxide concentration (labeled CO2) for 161 months. The two columns labeled CO2lag1 and CO2lag2 contain lag 1 carbon dioxide concentration and lag 2 carbon dioxide concentration. Lag 1 corresponds to values from the previous month and lag 2 corresponds to values from two months ago.
- Determine partial autocorrelations for the carbon dioxide concentration series. What do the results indicate about an auto-regression model for a time series that describes carbon dioxide concentration? That is, what are the “large” partial autocorrelations?
- Do a multiple regression with CO2 as the y-variable and CO2lag1 and CO2lag2 values as x-variables. Store the residuals.
- Write the estimated regression equation.
- Use the regression equation to find the fitted value for the carbon dioxide concentration for the 162nd month. [For this question you don’t need to use the method at the bottom of Section 14.3.]
- Plot the residuals versus order for the regression that you did in part (b). What is indicated about the validity of the independent errors assumption for the regression model?
- Obtain a PACF plot of the residuals for the regression that you did in part (b). What is indicated about the validity of independent errors assumption for the regression model?
- Are the conclusions in parts (c) and (d) in agreement? Suggest any remedial measures to remove the correlated nature of errors.
- (10×4 = 40 points) Use the “PennStateStudent” dataset. The data are from n = 900 students at Penn State, surveyed in February of 2007. We will use logistic regression to predict the probability that a student says they have ever cheated on a college exam. The y-variable is ChtdExam and possible responses are Yes and No.
- Do a logistic regression that relates the probability of having ever cheated on an exam to GPA. In Minitab, use Stat > Regression > Binary Logistic Regression > (Fit) Binary Logistic Model. Take the following actions:
- Enter ChtdExam as Response.
- Enter GPA in the Continuous Predictors box.
- Click Storage and request Fits (event probabilities). Click OK.
Write the sample logistic regression equation for this situation by filling in values for the coefficients b0 and b1 in the equation:
- Plot the stored predicted probabilities of having ever cheated on an exam versus GPA. [In Minitab, use Graph > Scatterplot and select “With Connect Line.” Use the column Fits1 as the y-variable and GPA as the x-variable.] Copy and paste the plot as part of your answer AND briefly describe what the plot shows about how the probability of having ever cheated on an exam is related to GPA.
- Examine the results that were generated in part (a) in order to find the odds ratio for GPA. Write a sentence that gives the value and interprets it in this situation. [If you’re using software that does not give the odds ratio, calculate it using the formula eb1.]
- Refer to the equation that you wrote in part (a). Use it to estimate the probability of having ever cheated on an exam for a student with a GPA = 3.0. (You can use the plot in part (b) to see if your answer is in the neighborhood of being correct.)
- Calculate the odds of having ever cheated on an exam for a student with a GPA = 4.0. [Hint: You can calculate this two different ways. Use the equation from part (a) and the fact that the odds of an event are p/(1–p). Alternatively, use the answers to parts (c) and (d) and the fact that the odds ratio multiplies the odds each time x is increased by one unit.]
- Now add the variable SkipClass as a predictor variable in the model (along with GPA). SkipClass is the number of classes student says he or she misses in a typical week. What is the evidence in the output that SkipClass is related to the probability of having ever cheated on an exam?
- For the model with two x-variables, write a sentence that gives the odds ratio for SkipClass and interprets it in the context of this situation.
- In the logistic model with two x variables, the odds of an event can be computed directly from the equation:
Use this equation to estimate the odds of having ever cheated on an exam for students with a 3.5 GPA who typically miss one class per week.
- In part (g) you wrote the odds ratio for SkipClass and in part (h) you found the odds of having ever cheated on an exam for students with a 3.5 GPA who typically skip one class per week. Use only the answers to parts (g) and (h) to determine the odds of having ever cheated on an exam for students with a 3.5 GPA who typically skip two classes per week.
- Four more variables in the dataset are the following:
- FakeID = whether student has ever used a fake id in order to be served alcohol.
- ChtdSO = whether student has ever cheated on another person with whom they were having a romantic relationship.
- SmokeCig = whether student smokes cigarettes.
- SmokeMJ = whether student has ever smoked marijuana.
All four of these variables are coded as either Yes or No. In the multiple logistic regression model for predicting the probability of ever having cheated on an exam, add the four variables just listed to the model along with GPA and SkipClass.
[In Minitab17, list these variables in the “Categorical Predictors.” In Minitab16, list these variables in the Model box and also in the Factors box. Minitab will create indicator variables for variables in the Factors box. In this case, each indicator variable created will equal 1 for Yes and 0 for No.]
Discuss the results with respect to these added variables only. Which variables are significant, which are not? Interpret the significant odds ratios.
- (4+7×3 = 25 points) Use the “Admit” dataset. A researcher is interested in how a number of variables affect admission into graduate school. The predictors are GRE scores, GPA, and prestige of the undergraduate institution (Rank), where GRE and GPA are quantitative but Rank is categorical with 4 levels (1 = highest prestige, 4 = lowest prestige). Rank could be treated as ordinal, but for this example we’ll assume it is nominal. The binary response variable here is Y = 1 if admitted and 0 if not.
- Fit a logistic regression model that relates the probability of being admitted to GRE scores, GPA, and Rank. [In Minitab, use Stat > Regression > Binary Logistic Regression > Fit Binary Logistic Model. Select Admit as the response, GRE and GPA as continuous predictors, and Rank as a categorical predictor. Before clicking OK, click “Results” and select “Expanded tables.”] Report the Coefficients table.
- What is the fitted equation? [Note that there is more than one way to write this, and whichever you prefer is fine as long as it is correct.]
- Interpret the estimated coefficient for GPA in the context of this problem.
- Interpret the estimated coefficient for Rank level 4 in the context of this problem. [Hint: Refer to the information about the Rank variable provided above. In your answer you’ll need to mention the reference level, which is Rank level 1.]
- Calculate and interpret the odds ratio for Rank level 4 vs. Rank level 3.
- Based on the model, find the odds that a student from a highest-ranked institution (rank 1), with a GPA of 3.8 and GRE score of 760, is admitted into graduate school.
- Find the probability of being admitted into graduate school for the same student as part (f).
- Use a Wald test to test if the coefficient for GRE is 0 using significance level α = 0.05.
The post STAT 501 – Homework 12 appeared first on My Assignment Online.