Assignment3

docx

School

Carleton University *

*We aren’t endorsed by this school

Course

1010

Subject

Statistics

Date

Apr 3, 2024

Type

docx

Pages

Uploaded by ChristyPaul

Assignment3_CHRPAU_643 1. Does there seem to be a problem with model bias? Steps - Open VS_BankLinear under Christy’s Folder - Duplicated into New page under Options - Added Partition as 50 percentage as New Data Item under Data pane - Add partition as Partition ID under roles - Click Maximize then Assessment tab. Double click the percentile to sort from low to high. Assessment pane in linear regressions shows the plot of average predicted and average response values against binned data. Significant differences in both variables’ values can indicate bias. The above assessment plot shows a significant difference between 10 th and 25 th percentile and after 90 th percentile. This is in comparison to predicted average and observed average values of the model. This shows us there is some bias in the model. 2. Do the residuals seem to satisfy the assumption of constant variance?

Steps: - Select Residual Statistics tab to view cluster and outliers better As the figure indicates above, there are some extreme outliers to the far right at predicted value 24000. Meanwhile there is a clear cluster between 6000 and 18000, a pattern which is violating the model. Model is also showing a constant variance around the 0 mean value and homoscedasticity, showing similar variances. As assumption of constant variance, the relative spread of the residual values changes as the predicted values change. Thus, Constant Variance is not violated. 3. Does the GLM model solve the problems that you found in the linear model, if any?

Steps - Holding down ALT, duplicate linear model into a new page as GLM - Set parameters by selecting Informative missingness check box, set a Backward elimination method under Significance level selection criterion and default significance level of .01 and link function to Log . - Select AIC under GLM menu. Change Linear model to AIC as well for comparison. AIC surged from 1828796.04 to 2018834.95 where there is a small increase on AIC value for GLM. Smaller AIC indicates better model. While the initial model has little problems, a GLM model did not seem to improve it. And, when assessment pane is observed, bias on the right and left tails has slight increases compared to Linear regression model. In residual plot, there is still homoscedasticity and with outlier issues observed. Thus, GLM do not solve the problems found in the linear model mentioned in answer to Question 2. 4. d. What main effects were removed from this model during the backward variable selection process? (Do not include the _miss variables.) Steps: - Maximize view of GLM

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version

Access to all documents
Unlimited textbook solutions
24/7 expert homework help

- Select Selection Summary. Steps that are not zero are the effects removed. Thus, Step 1, 2, 3, and 4. Effects removed include: - Customer Tenure - Count Prchsd Past 3 Years Dir Promo Resp - Avg Sales Past 3 Years Dir Promo Resp - Average Sales Past 3 Years