Read the Takyo Software Cataloger case in the Shmueli textbook (pg. 482) and use the corresponding spreadsheet Takyo.xls is available inModule 0: Spreadsheet Assignment Data(Link will open in new...

The data set contains main numerical variables like number of transactions in last year at source catalog, number of days since the last and first update of the customer and some categorical binary variables like web order, gender of the customer, address of the customer based on residence, whether the customer purchased in test mailing. All these variables are taken as predictor variables to create the multiple linear regression model and the variable, spending, which describes the amount of spending by a customer in test mailing, has been taken as the dependent variable or the response variable. Since in the data set, Purchase variable has taken the value 1 only, we have neglected that variable and created the linear model based on the other variables. The first variable has been created including all the variables at once and then the backward method has been applied to remove the variables with worst insignificant p-values and the process is repeated until a model is found with all significant variables. The final model is thus shown along with the regression equation and the output.
    SUMMARY OUTPUT
    
    
    
    
    
    
    
    
    
    
    
    
    
    Regression Statistics
    
    
    
    
    
    Multiple R
    0.669722
    
    
    
    
    
    R Square
    0.448528
    
    
    
    
    
    Adjusted R Square
    0.446867
    
    
    
    
    
    Standard Error
    164.1943
    
    
    
    
    
    Observations
    1000
    
    
    
    
    
    
    
    
    
    
    
    
    ANOVA
    
    
    
    
    
    
     
    df
    SS
    MS
    F
    Significance F
    
    Regression
    3
    21839438.52
    7279813
    270.025
    3.2105E-128
    
    Residual
    996
    26851937.81
    26959.78
    
    
    
    Total
    999
    48691376.33
     
     
     
    
    
    
    
    
    
    
    
     
    Coefficients
    Standard Error
    t Stat
    P-value
    Lower 95%
    Upper 95%
    Intercept
    81.73259
    14.93718327
    5.471754
    5.64E-08
    52.42063041
    111.0446
    Freq
    87.85543
    3.466726766
    25.34247
    9.5E-110
    81.0525005
    94.65835
    last_update_days_ago
    -0.02036
    0.004920589
    -4.13805
    3.8E-05
    -0.030017573
    -0.01071
    Address_is_res
    -84.7679
    12.64848479
    -6.70182
    3.44E-11
    -109.5886598
    -59.9472
From the regression output it is evident that the overall model is highly significant with a significant p-value of the F-statistic. Which implies that this model is better than the null model and hence can be accepted for the prediction purpose with the aforementioned variables finally selected by the stepwise backward regression method. The predictor variables seem to be highly significant along with the intercept. The coefficient of determination is found to be 0.446867 which implies that almost 45% of the variance in the response...