Week 3: Basic regression - 4. How useful is a linear model
Transcript of Week 3: Basic regression - 4. How useful is a linear model
![Page 1: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/1.jpg)
Week 3: Basic regression
4. How useful is a linear model
Stat 140 - 04
Mount Holyoke College
Dr. Shan Shan Slides posted at http://sshanshans.github.io/stat140
![Page 2: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/2.jpg)
2020 U.S. Election Example
1
![Page 3: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/3.jpg)
2020 U.S. Election Example
2
![Page 4: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/4.jpg)
What does the intercept mean here?
Is it useful?
What is R2?
3
![Page 5: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/5.jpg)
What does the intercept mean here?
Is it useful?
What is R2?
3
![Page 6: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/6.jpg)
Statistics Exam Example
The two scatterplots below show the relationship between finaland mid-semester exam grades recorded during several years fora Statistics course at a university.
I Final exam the final
I Exam 1 first midterm
I Exam 2 second midterm
4
![Page 7: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/7.jpg)
Poll question
Which of these models would you prefer to use for predictingsales?
a Exam 1
b Exam 2
5
![Page 8: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/8.jpg)
Think together
Being as specific and concrete as possible, write down a rule forselecting your preferred model
1. based only on visual characteristics of the plot.
2. based only on a quantitative summary of the data. Youcan describe how you would calculate your numericsummary of the data in a general sense; if you’d like youcan write down a formula.
6
![Page 9: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/9.jpg)
Variation of residuals
Residuals:
I ei = yi − yi (vertical distance between point and line)I Smaller residuals mean the predictions were better.I The key is to measure the spread of residuals.
7
![Page 10: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/10.jpg)
Residual standard error
Measure spread of residuals with the standard deviation. We callthis the residual standard error, sRES.
I Exam 1: 4.28
I Exam 2: 3.26
8
![Page 11: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/11.jpg)
Variation accounted by the model
The variability in the residuals describes how much variationremains after using the model
9
![Page 12: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/12.jpg)
Variation accounted by the model
Let’s compute the reduction in variation.
s2sales − s2RESs2sales
= 0.61
This number describes the amount of variation in the y-variablethat is explained by the least squares line.
An value of 61% indicates that 61% of the variation in finalexam grades can be accounted for by Exam 1 grades.
10
![Page 13: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/13.jpg)
More practice
Variation accounted by the model
I Exam 1: 0.61
I Exam 2: 0.73
meaning,
I 61% of the variation in final exam grades can be accountedfor by Exam 1 grades;
I 73% of the variation in final exam grades can be accountedfor by Exam 2 grades
11
![Page 14: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/14.jpg)
How do we compute the reduction?
Statisticians found the variation accounted by the model can becomputed by R2, the square of correlation.
Square of the correlation coefficient R: between 0 and 1, closerto 1 is better.
R2 describes the amount of variation in the y-variable that isexplained by the least squares line.
12
![Page 15: Week 3: Basic regression - 4. How useful is a linear model](https://reader031.fdocuments.in/reader031/viewer/2022030113/621cc12fbf285c171a41f2bd/html5/thumbnails/15.jpg)
Compute R2 from R
linear fit ← lm(Mortality ∼ Calcium, data = mortality water)summary(linear fit)
13