Scribd
Upload
262 views6 pages

MY NOTES Linear Regression

This document contains notes from an AP Statistics class on linear regression. It discusses using a linear regression model to find the relationship between two variables, the protein and fat content of Burger King menu items. The model finds the slope and y-intercept of the regression line. The residual plot is examined to check if the linear model is a good fit to the data. The R-squared value indicates that 69% of the variability in fat content is explained by the linear relationship to protein content. TI calculator instructions are provided for performing linear regression analysis.

Uploaded by

bveshpsu
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
262 views6 pages

MY NOTES Linear Regression

This document contains notes from an AP Statistics class on linear regression. It discusses using a linear regression model to find the relationship between two variables, the protein and fat content of Burger King menu items. The model finds the slope and y-intercept of the regression line. The residual plot is examined to check if the linear model is a good fit to the data. The R-squared value indicates that 69% of the variability in fat content is explained by the linear relationship to protein content. TI calculator instructions are provided for performing linear regression analysis.

Uploaded by

bveshpsu
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 6

AP Stats – Mr.

Veshio
Unit 2 – Chapter 8
**ActivStats** 8-1 “Manatees and Motorboats”

Let’s look at a scatterplot of these data on ActivStats…

How could we come up with an equation of this line? What will this
help us do?

I -Residuals
**Using the attached data on Burger King menu items, make a
scatterplot of the data on your TI**

**ActivStats** 8-3 – “Residuals”

When a “line of best fit” is drawn, we refer to it as a LINEAR MODEL


 Use this line to make ESTIMATES

Predicted Values: Estimates made from a model, called “y-hat”

Residual: Difference between the observed value (y) and its


associated predicted value (y-hat)
 Tells us how far off the model’s prediction is at that point

Residual = Observed – Predicted

Line of best fit – couldn’t we just add up the residuals, and see when
the sum is minimized?
 NO! – Why? The positive and negative residuals would cancel
each other out (LIKE STANDARD DEVIATION)
 How did we fix it? SQUARE THE RESIDUALS

Least Squares Line (aka Line of Best Fit) – minimizes the sum of the
squares of the residuals
 This line, and ONLY this line, will have the smallest variation
from the data as seen in the residuals

II – Correlation and the Least Squares Line

Let’s standardize the data – translate them into z-scores:

TI TIPS
STAT -> 1: Edit -> To create a new list, hit 2ND-> INS (above DEL
button)
Name it ZFAT -> Move your cursor on the name and hit ENTER -> The
cursor will be at the bottom now

AP Stats – Mr. Veshio -1-


Unit 2 – Chapter 8
AP Stats – Mr. Veshio
Unit 2 – Chapter 8
Hit 2ND -> LIST (above STAT button) -> Select FAT
You can now use this as a variable – whatever you tell the calculator to
do with this list, it will do it for every one of the values in list FAT -> To
convert to z-scores, we need to subtract the mean of the data, and
divide by the standard deviation

Run 1-VarStats on FAT and PROT to find the mean and standard
deviation

(FAT – 23.11) / 16.29 -> Hit Enter -> The calc will now fill ZFAT with the
z-scores for all 32 items’ fat content

Do the same for PROT

(PROT – 17.09) / 13.79

If you construct scatterplot, you will see a picture similar to Fig. 8.2 on
p. 170 minus the line – we don’t know how to draw it yet – but do we
know anything about it?

 Should go through the point (x , y), the averages of both


variables – but on a standardized scatterplot, that point is (0 , 0)
 Lines that go through the origin look like this: y = mx, where
“m” is the slope – but since variables were standardized, the
equation looks a little different:
zy = mzx
 Which slope will be “best” i.e. which slope gives us the line that
minimizes the sum of the residuals? THE CORRELATION
COEFFICIENT
zy = rzx
 “Moving one standard deviation away from the mean in x moves
our ESTIMATE r standard deviations away from the mean in y”

AP Stats – Mr. Veshio -2-


Unit 2 – Chapter 8
AP Stats – Mr. Veshio
Unit 2 – Chapter 8

Fig. 8.3 – The slope of a regression line on a standardized scatterplot


Just Checking
1.) about 0.85 standard deviations
2.) about 1.7 standard deviations below

***FROM MATH BOX, p. 172-173***


 Slope of line of best fit for z-scores is “r”
 Slope of regression line is b = rsy/sx
 1 – r2 = % of variability NOT explained by the regression line, so
 r2 = % of variability in y that IS explained by x
o explained by our model -> at most 100% (which is why r is
between -1 and 1)

III – Specifying the Regression Line

Instead of y = mx + b, we’ll use

y = b0 + b1x

You’ll need to find…


b0 = slope
b1 = y-intercept

MINI STEP-BY-STEP

1.) Find Slope

b1 = rsy/sx = (0.83 x 16.4 g fat) / 14g protein = 0.97g fat per g protein

2.) Find y-intercept

 Regression line will always pass through (x , y) -> substitute in


with the slope, and solve:

x = 17.2g protein
y = 23.5g fat

23.5 g fat = b0 + 0.97g fat/g protein · 17.2g fat


b0 = 6.8g fat

3.) Put slope and y-intercept back in, using specific variables

Fat = 6.8 + 0.97Protein

AP Stats – Mr. Veshio -3-


Unit 2 – Chapter 8
AP Stats – Mr. Veshio
Unit 2 – Chapter 8

How to Interpret

Slope: “y units per x” -> one additional gram of protein associated


with additional 0.97g of fat
 UNITS MATTER – include them

Intercept: value of the line when x=0 -> 6.8g fat for an item with no
protein, on average
 Zero is not usually plausible value -> y-intercept only serves as a
starting point

***SINCE CORRELATION IS AT THE HEART OF REGRESSION, YOU MUST


CHECK YOUR THREE CONDITIONS FIRST***

Just Checking p. 175

3.) Price increases on average $122,740 for every 1000 square feet

4.) dollars / square foot

5.) $245,480

6.) $377,784

IV – Residuals Revisited

Residuals: part of the data that HASN’T been modeled

Residual = Data – Model

e=y-y

“What did the model miss?” -> If our model is “good”, nothing
interesting should be left behind -> PLOT THE RESIDUALS

AP Stats – Mr. Veshio -4-


Unit 2 – Chapter 8
AP Stats – Mr. Veshio
Unit 2 – Chapter 8

 What should we see? NOTHING! No patterns, curves, bends – it


should stretch horizontally, with the same amount of scatter
throughout
 Residual plot is a good way to check “Straight Enough Condition”
o Plot them against either variable, usually y

V – R2 – Variation Accounted For

r2 = squared correlation

 Represents the fraction of data’s variation accounted for by the


model
 REMEMBER – our model isn’t perfect – some of the variation
WON’T be accounted for -> 1 - r2 of the variation is left in the
RESIDUALS

 Denoted R2 -> given as a % -> WHY?

Burger King Example

Variation in fat = variance = s2 =16.42 = 268.42

Variation in Residuals = 83.195

Variance of Resid. / Variance of Fat = 83.195 / 268.42 = 0.31 or 31%

 This 31% is the variation left in the residuals -> to find the
variation explained by the model (the rest of it), take 100% -
31% = 69%

HOW TO INTERPRET R2
“According to our linear model, 69% of the variability in fat content of
30 Burger King items is accounted for by variation in protein content”

 ALWAYS report it with a scatterplot and linear regression model

AP Stats – Mr. Veshio -5-


Unit 2 – Chapter 8
AP Stats – Mr. Veshio
Unit 2 – Chapter 8

TI TIPS – LINEAR REGRESSION

STAT -> Calc -> 8: LinReg (a + bx)

Enter YR, TUIT, Y1 -> Y1 can be found by VARS -> y-Vars -> 1:Function
 Will create a regression equation and store into Y1

**CHECK RESIDUALS**
 Calcs make a residual list when it runs LinReg -> List name is
always RESID -> if you don’t see it them in the lists, create a new
list and store them in it by finding RESID in the list of list names
 STAT PLOT -> X: YR Y: RESID -> ZOOM -> 9: ZoomStat (might
help to turn axes off)
 WHAT DO YOU SEE? AND WHAT DOES THIS MEAN?

What Can Go Wrong?


 Linearity is KEY
 Don’t fit a line to a non-linear relationship
 Re-express if possible
 Watch out for extreme values – OUTLIERS!!
 Causation is STILL not valid
 Fitting a regression line doesn’t help your case
 R does not work by itself
2

 Always look at the scatterplot in conjunction

HW: re – read Ch. 8, do #’s 2, 4, 15, 24, 40, 42

AP Stats – Mr. Veshio -6-


Unit 2 – Chapter 8

You might also like