- 60 Views
- Uploaded on
- Presentation posted in: General

Homework Assignment

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Homework Assignment

10.17, 10.19, 10.23, 10.29, 10.32

Due in Class Dec 1

Last Time:Finished Contingency TablesReviewed Basics on Linear Regression

- How many observations do I expect to get in cell (i,j)?
- If the Null Hypothesis holds, i.e., if the columns and rows are independent, then I expect the number of observations in cell (i,j) to be

How to compare??

Both are unknown!

How to compare??

Both are unknown!

Table F (T-20)

Example: Development in 1st grade

Example: Development in 1st grade

FIXED

Homogeneity of

parallel samples

255

32

FIXED

Equivalent Equations

The rest is the same as in previous scenario,

i.e., we get the same Chi-square again.

Today:From descriptive to inference statistics…Estimation and Hypothesis Testingfor Linear Regression

Estimation: (Confidence Intervals)

Point

estimate

critical

value

Std. dev. of

point estimate

±

·

For instance: Confidence Interval for the mean:

±

·

Hypothesis Testing:

1pt

2pts

3pts

“Popularity Score” = Average Score

“Social Competence Score”

“Popularity Score”

“Social Competence Score”

Example:

Children X: Popularity,

Y: Social competence

Goal: Explain (linear) relationship between X and Y

Y

1

X

Explain the (linear) relationship (if it exists)

between random variable X and random variable Y.

Four assumptions

about the error term

4

For different

values of X,

the error terms

are uncorrelated

1

3

2

The error term

is a normally

distributed

random variable

No matter what

value X takes,

the error has

a mean of zero

Error

Second assumption: Average error is zero for each value of X

1

X

3rd assumption: Error has same standard deviation for each value of X

Y

Error

X

Y

X

Y

X

We will sample data to estimate the parameters.

This leads to point estimates, confidence intervals

and hypothesis testing for each parameter,

in addition to a general test of the model as a whole.

Degrees of freedom

loose 1 df for X

loose 1 df for Y

Sampling Distributions

Don’t

Know!

Hypothesis Testing

Remember the general rule for Confidence Intervals:

Point

estimate

critical

value

Std. dev. of

point estimate

±

·

Point

estimate

critical

value

Std. dev. of

point estimate

±

·

Hypothesis Test on Slope

If p-value of the standardized statistic < then reject H0

and conclude that there is indeed a linear relationship

Hypothesis Test on Slope

p-value

< .001

How much

y differs

from mean

How much

predicted y

differs

from mean

residual

/

error

Involves

only data

Sum Squares

Total

(SST)

Sum Squares

Error

(SSE)

Sum Squares

Model

(SSM)

Variability

in the

Data

Variability

unaccounted

for

Variability

accounted for

by the Model

Degrees of Freedom

DFT = n-1

Degrees of Freedom

DFE=n-2

Degrees of Freedom

DFM = 1

=

+

Degrees of Freedom

DFT = n-1

Degrees of Freedom

DFE=n-2

Degrees of Freedom

DFM = 1

Mean Squares

Total

(MST)

Mean Squares

Error

(MSE)

Mean Squares

Model

(MSM)

Mean Squares

Error

(MSE)

Mean Squares

Error

(MSE)

It can be shown that

the Null Hypothesis

implies that

MSM is also an unbiased

estimator of

Table E

df in the numerator

df in the denominator

p

p-value here: <.001

The Square of a t Random Variable with n-2 degrees of freedomis an F Random Variablewith 1 degree of freedom in the numerator andwith n-2 degrees of freedom in thedenominator.

- Treated the X variable as if it were fixed.
- Now, let’s think of X and Y as both being random variables (jointly distributed).
- Let’s assume they are both normally distributed (i.e. they are “jointly normal”)
- We can define a population correlation ρ
- We can use the sample correlation r as an estimate of ρ.

Hypothesis Test about Correlation

Question: Can we / can we not draw a line close to the data?

Answer: No, unless we provide sufficient evidence that we can.

loose 1 df for X

loose 1 df for Y

If p-value of the standardized statistic < then reject H0

and conclude that there is indeed a linear relationship

In our Example:

Percent of Variance

explained by the model