lord william thomson, 1st baron kelvin

45

Upload: eve

Post on 24-Feb-2016

82 views

Category:

Documents


0 download

DESCRIPTION

“ I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it ”. Lord William Thomson, 1st Baron Kelvin. Statistics =. “getting meaning from data”. (Michael Starbird ). measures of central values, measures of variation, - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Lord William Thomson, 1st Baron  Kelvin
Page 2: Lord William Thomson, 1st Baron  Kelvin

“I often say that when you can measure what you are

speaking about, and express it in numbers, you know something about it”Lord William Thomson,

1st Baron Kelvin

Page 3: Lord William Thomson, 1st Baron  Kelvin
Page 4: Lord William Thomson, 1st Baron  Kelvin

Statistics =“getting meaning

from data”(Michael Starbird)

Page 5: Lord William Thomson, 1st Baron  Kelvin

descriptivestatistics

“inferential”statistics

measures of central values,measures of variation,

visualization

beatingchance!

Page 6: Lord William Thomson, 1st Baron  Kelvin

“inferential”statistics

beatingchance!

Page 7: Lord William Thomson, 1st Baron  Kelvin

“inferential”statistics

beatingchance!

SamplePopulation

inference

PARAMETERS

ESTIMATES

Page 8: Lord William Thomson, 1st Baron  Kelvin

But what’s the valueof inferential statisticsin our field??1. More explicit theories

2. More constraints on theory

3. (Limited) generalizability

Page 9: Lord William Thomson, 1st Baron  Kelvin

H0 = there is no difference, or there is no correlation

Ha = there is a difference; there is a correlation

The (twisted) logic of hypothesis testing

Page 10: Lord William Thomson, 1st Baron  Kelvin

Type I error =behind bars…… but not guilty

Type II error =guilty…… but not

behind bars

The (twisted) logic of hypothesis testing

Page 11: Lord William Thomson, 1st Baron  Kelvin

p < 0.05What does

it really mean?

Page 12: Lord William Thomson, 1st Baron  Kelvin

p < 0.05= Given that H0 is true,

this data would befairly unlikely

Page 13: Lord William Thomson, 1st Baron  Kelvin

One-sample t-test

Unpairedt-test ANOVA

ANCOVA Regression

MANOVAχ2

test

Discrimant

Function Analysis

Pairedt-test

Page 14: Lord William Thomson, 1st Baron  Kelvin

One-sample t-test

Unpairedt-test ANOVA

ANCOVA Regression

MANOVAχ2

test

Discrimant

Function Analysis

Pairedt-test

Page 15: Lord William Thomson, 1st Baron  Kelvin

Linear Model

Page 16: Lord William Thomson, 1st Baron  Kelvin

GeneralLinear Model

Page 17: Lord William Thomson, 1st Baron  Kelvin

GeneralLinear Model

GeneralizedLinear Model

GeneralizedLinearMixed Model

Page 18: Lord William Thomson, 1st Baron  Kelvin

GeneralLinear Model

GeneralizedLinear Model

GeneralizedLinearMixed Model

Page 19: Lord William Thomson, 1st Baron  Kelvin

what you measure

what you manipulate

“response”

“predictor”

RT ~ Noise

Page 20: Lord William Thomson, 1st Baron  Kelvin
Page 21: Lord William Thomson, 1st Baron  Kelvin
Page 22: Lord William Thomson, 1st Baron  Kelvin

best fitting line(least squares estimate)

Page 23: Lord William Thomson, 1st Baron  Kelvin

the intercept

the slope

Page 24: Lord William Thomson, 1st Baron  Kelvin

Same intercept, different slopes

Page 25: Lord William Thomson, 1st Baron  Kelvin

Positive vs. negative slope

Page 26: Lord William Thomson, 1st Baron  Kelvin

Same slope, different intercepts

Page 27: Lord William Thomson, 1st Baron  Kelvin

Different slopes and intercepts

Page 28: Lord William Thomson, 1st Baron  Kelvin

The Linear Model response ~ intercept + slope * predictor

Page 29: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ b0 + b1*X1

coefficients

Page 30: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ b0 + b1*X1

slopeintercept

Page 31: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ 300 + 9*X1

slopeintercept

Page 32: Lord William Thomson, 1st Baron  Kelvin

With Y ~ 300 + 9 *x,what is the response time for a

noise level of x = 10?

30010

300 + 9*10 = 390

Page 33: Lord William Thomson, 1st Baron  Kelvin

Deviation from regression line

= residual

“fitted values”

Page 34: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ b0 + b1*X1 + error

Page 35: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ b0 + b1*X1 + error

Page 36: Lord William Thomson, 1st Baron  Kelvin

is continuous

is continuous,too!

Page 37: Lord William Thomson, 1st Baron  Kelvin

RT ~ Noise

men

women

Page 38: Lord William Thomson, 1st Baron  Kelvin

men

women

RT ~ Noise + Gender

Page 39: Lord William Thomson, 1st Baron  Kelvin

The Linear ModelY ~ b0 + b1*X1 + b2*X2

coefficientsof slopes

coefficient ofintercept

noise(continuous)

gender(categorical)

Page 40: Lord William Thomson, 1st Baron  Kelvin

The Linear Model

“Response” ~ Predictor(s)

Has to be onething

Can be one thingor many things

“multiple regression”

Page 41: Lord William Thomson, 1st Baron  Kelvin

The Linear Model

“Response” ~ Predictor(s)

(we’ll relaxthat constraint

later)

Can be of any data type

(continuous or categorical)

Has to becontinuous

Page 42: Lord William Thomson, 1st Baron  Kelvin

The Linear Model

RT ~ noise + gender

examples

pitch ~ polite vs. informal

Word Length ~ Word Frequency

Page 43: Lord William Thomson, 1st Baron  Kelvin

Edwards & Lambert (2007); Bohrnstedt & Carter (1971); Duncan (1975); Heise (1969); in Edwards & Lambert (2007)

Correlation is (still) not causation

Page 44: Lord William Thomson, 1st Baron  Kelvin

“Response” ~ Predictor(s)

Assumed directionof causality

Correlation is (still) not causation

Page 45: Lord William Thomson, 1st Baron  Kelvin