Homework 2
数据挖掘作业辅导 Please limit your answer to Q2 to 8 pages, organized into a coherent typed data analysis report. Answers to Q1 may be either typed or ···
数据挖掘作业辅导
Please limit your answer to Q2 to 8 pages, organized into a coherent typed data analysis report. Answers to Q1 may be either typed or handwritten.
1.Suppose you are trying to predict starting salary in a certain position (y, in thousands of dollars) from undergraduate GPA (x1, from 0 to 4), gender (with only two options recorded, x2 = 1 if female, 0 if male), and relevant prior experience (x3, in years). Suppose you fifit a linear model
y = β0 + β1x1 + β2x2 + β3x3 + β4x1x2 + β5x1x3
and obtained the OLS estimate βˆ = (40, 3, −2, 1.5, −0.5, −0.1)T . [50 points]
(a) Predict the starting salary for a female employee with a GPA of 3.5 and 2 years experience. [5 points]
(b) For a given value of GPA and years of experience, do male or female employees earn more, on average? By how much? Show the work you did to obtain your answer to each of these two questions, or explain why it is not possible to answer. [10 points]
(c) The coefficient for the interaction between GPA and gender is negative (βˆ4 = −0.5). What is the interpretation of that? [10points]
(d) The smallest (in absolute value) coefficient is βˆ5 = −0.1. Does that mean that there is no interaction between GPA and experience? Explain. [10 points]
(e) For someone with no prior experience (x3 = 0), sketch the two lines showing the relationship between income and GPA for males and females. Write down the equations for both lines. [10 points]
(f) What hypothesis would you need to test to decide whether the population lines in the previous question should be parallel? [5 points]
数据挖掘作业辅导
2.This exercise relates to the Carseats data set in the ISLR package. You may use help(Carseats) to learn more about the data set. [50 points]
(a) Fit a multiple regression model to predict Sales using all other variables in the model with no interactions (the full model) Report the values of coeffiffifficients, and how well the model fifits (using R2 ). Include diagnostic residual plots and comment on any in teresting features. [10 points]
(b) Which variables have signifificant p-values? What is the hypothesis corresponding to the p-value which appears in the summary table for the variable Urban? [6 points]
(c) Drop all the variables that are not signifificant in the full model (Note: this is not the best way to do model selection; we will study better ways later). Fit the linear model with the remaining variables and no interactions (the reduced model). It will
include one categorical variable, ShelveLoc. Compare the fifit of the reduced model to the fifit of the full model using R2 . [5 points]
(d) Use the anova() command to formally compare the full and reduced models and state your conclusion. Comment on the diffffer ence between their R2 in light of your conclusion. [6 points]
(e) Write out the reduced model in equation form and interpret the coeffiffifficients. Be careful with the coeffiffifficients of the categorical variable. [6 points]
(f) Add an interaction term between the categorical variable ShelveLoc and the variable Price to the reduced model. Report the estimated coeffiffifficients, and interpret the coeffiffifficients of the interaction term. Do the corresponding p-values suggest the interaction term is necessary? [10 points]
(g) Use anova() to formally test whether the interaction term is needed, and state your conclusion. [7 points]
更多代写：程式代寫價格 siele代考 案例分析 美国essay格式 HIS代写 代写java编程作业
