Slide 1

Slide 1 text

Statistical Thinking for Data Science Chris Fonnesbeck Vanderbilt University

Slide 2

Slide 2 text

No content

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

21/22 falling 7+ stories survived

Slide 5

Slide 5 text

2 fell together

Slide 6

Slide 6 text

40% at night

Slide 7

Slide 7 text

“Even more surprising, the longer the fall, the greater the chance of survival.”

Slide 8

Slide 8 text

2 to 32 stories (average = 5.5)

Slide 9

Slide 9 text

?

Slide 10

Slide 10 text

"... 132 such victims were admitted to the Animal Medical Center on 62nd Street in Manhattan ..."

Slide 11

Slide 11 text

"Found" Data

Slide 12

Slide 12 text

convenience sample

Slide 13

Slide 13 text

Missing Data

Slide 14

Slide 14 text

Representative

Slide 15

Slide 15 text

Statistical Issues

Slide 16

Slide 16 text

Big Data

Slide 17

Slide 17 text

“With enough data, the numbers speak for themselves ” Chris Anderson, Wired

Slide 18

Slide 18 text

Alfred Landon

Slide 19

Slide 19 text

Literary Digest Straw Poll

Slide 20

Slide 20 text

"Next week, the first answers from these ten million will begin the incoming tide of marked ballots, to be triple-checked, verified, five-times cross-classified and totalled."

Slide 21

Slide 21 text

2.4 million returns

Slide 22

Slide 22 text

41 - 55

Slide 23

Slide 23 text

No content

Slide 24

Slide 24 text

George Gallup

Slide 25

Slide 25 text

Sampled 50,000

Slide 26

Slide 26 text

66%

Slide 27

Slide 27 text

Random Sampling

Slide 28

Slide 28 text

No content

Slide 29

Slide 29 text

Bias

Slide 30

Slide 30 text

No content

Slide 31

Slide 31 text

No content

Slide 32

Slide 32 text

Self-selection Bias

Slide 33

Slide 33 text

No content

Slide 34

Slide 34 text

For some estimate of unknown quantity ,

Slide 35

Slide 35 text

p = 0.5 sample_sizes = [10, 100, 1000, 10000, 100000] replicates = 1000 biases = [] for n in sample_sizes: bias = np.empty(replicates) for i in range(replicates): true_sample = np.random.normal(size=n) negative_values = true_sample<0 missing = np.random.binomial(1, p, n).astype(bool) observed_sample = true_sample[~(negative_values & missing)] bias[i] = observed_sample.mean() biases.append(bias)

Slide 36

Slide 36 text

No content

Slide 37

Slide 37 text

Accuracy Mean Squared Error

Slide 38

Slide 38 text

“The numbers have no way of speaking for themselves” Nate Silver

Slide 39

Slide 39 text

White House Big Data Partners Workshop

Slide 40

Slide 40 text

White House Big Data Partners Workshop 19 Participants 0 Statisticians

Slide 41

Slide 41 text

NSF Working Group on Big Data

Slide 42

Slide 42 text

NSF Working Group on Big Data 100 experts convened 0 statisticians

Slide 43

Slide 43 text

Moore Foundation Data Science Environments

Slide 44

Slide 44 text

Moore Foundation Data Science Environments 0 directors with statistical expertise

Slide 45

Slide 45 text

NIH BD2K Executive Committee

Slide 46

Slide 46 text

NIH BD2K Executive Committee 17 committee members 0 statisticians

Slide 47

Slide 47 text

Feeling left out?

Slide 48

Slide 48 text

It's our own fault

Slide 49

Slide 49 text

“Almost everything you learned in your college statistics course was wrong”

Slide 50

Slide 50 text

Typical introductory statistics syllabus 1.Descriptive statistics and plotting

Slide 51

Slide 51 text

Typical introductory statistics syllabus 1.Descriptive statistics and plotting 2.Basic probability

Slide 52

Slide 52 text

Typical introductory statistics syllabus 1.Descriptive statistics and plotting 2.Basic probability 3.Hypothesis testing

Slide 53

Slide 53 text

Typical introductory statistics syllabus 1.Descriptive statistics and plotting 2.Basic probability 3.Hypothesis testing 4.Experimental design

Slide 54

Slide 54 text

Typical introductory statistics syllabus 1.Descriptive statistics and plotting 2.Basic probability 3.Hypothesis testing 4.Experimental design 5.ANOVA

Slide 55

Slide 55 text

Statistical Hypothesis Testing

Slide 56

Slide 56 text

No content

Slide 57

Slide 57 text

No content

Slide 58

Slide 58 text

Test Statistic

Slide 59

Slide 59 text

T-statistic

Slide 60

Slide 60 text

No content

Slide 61

Slide 61 text

No content

Slide 62

Slide 62 text

No content

Slide 63

Slide 63 text

p-value

Slide 64

Slide 64 text

No content

Slide 65

Slide 65 text

No content

Slide 66

Slide 66 text

false positive rate

Slide 67

Slide 67 text

"The value for which , or 1 in 20, is 1.96 or nearly 2; it is convenient to take this point as a limit in judging whether a deviation ought to be considered significant or not." R.A. Fisher

Slide 68

Slide 68 text

p-value

Slide 69

Slide 69 text

the probability that the observed differences are due to chance

Slide 70

Slide 70 text

the probability that the observed differences are due to chance

Slide 71

Slide 71 text

a measure of the reliability of the result

Slide 72

Slide 72 text

a measure of the reliability of the result

Slide 73

Slide 73 text

the probability that the null hypothesis is true

Slide 74

Slide 74 text

the probability that the null hypothesis is true

Slide 75

Slide 75 text

"If an experiment were repeated infinitely, p represents the proportion of values more extreme than the observed value, given that the null hypothesis is true."

Slide 76

Slide 76 text

H0 : Mean duckling body mass did not differ among years.

Slide 77

Slide 77 text

H0 : Mean duckling body mass did not differ among years.

Slide 78

Slide 78 text

H0 : The prevalence of autism spectrum disorder for males and females were equal.

Slide 79

Slide 79 text

H0 : The prevalence of autism spectrum disorder for males and females were equal.

Slide 80

Slide 80 text

H0 : The density of large trees in logged and unlogged forest stands were equal

Slide 81

Slide 81 text

H0 : The density of large trees in logged and unlogged forest stands were equal

Slide 82

Slide 82 text

Statistical Straw Man

Slide 83

Slide 83 text

Statistical hypotheses are not interesting

Slide 84

Slide 84 text

Hypothesis tests are not decision support tools

Slide 85

Slide 85 text

Multiple Comparisons

Slide 86

Slide 86 text

No content

Slide 87

Slide 87 text

Family-wise Error Rate >>> 1. - (1. - 0.05) ** 20 0.6415140775914581

Slide 88

Slide 88 text

import seaborn as sb import pandas as pd n = 20 r = 36 df = pd.concat([pd.DataFrame({'y':np.random.normal(size=n), 'x':np.random.random(n), 'replicate':[i]*n}) for i in range(r)]) sb.lmplot('x', 'y', df, col='replicate', col_wrap=6)

Slide 89

Slide 89 text

No content

Slide 90

Slide 90 text

Statistically Significant!

Slide 91

Slide 91 text

No content

Slide 92

Slide 92 text

"Despite a large statistical literature for multiple testing corrections, usually it is impossible to decipher how much data dredging by the reporting authors or other research teams has preceded a reported research finding."

Slide 93

Slide 93 text

What's the Alternative?

Slide 94

Slide 94 text

Build models and use them to estimate things we care about

Slide 95

Slide 95 text

Effect size estimation

Slide 96

Slide 96 text

Data-generating Model

Slide 97

Slide 97 text

No content

Slide 98

Slide 98 text

No content

Slide 99

Slide 99 text

Florida manatee Trichechus manatus

Slide 100

Slide 100 text

No content

Slide 101

Slide 101 text

No content

Slide 102

Slide 102 text

No content

Slide 103

Slide 103 text

occupied?

Slide 104

Slide 104 text

occupied? available?

Slide 105

Slide 105 text

occupied? available? seen?

Slide 106

Slide 106 text

No content

Slide 107

Slide 107 text

Estimating visibility

Slide 108

Slide 108 text

No content

Slide 109

Slide 109 text

No content

Slide 110

Slide 110 text

No content

Slide 111

Slide 111 text

No content

Slide 112

Slide 112 text

No content

Slide 113

Slide 113 text

No content

Slide 114

Slide 114 text

No content

Slide 115

Slide 115 text

Bayesian Statistics

Slide 116

Slide 116 text

No content

Slide 117

Slide 117 text

No content

Slide 118

Slide 118 text

Bayes' Formula

Slide 119

Slide 119 text

Probabilistic Modeling

Slide 120

Slide 120 text

Evidence-based Medicine

Slide 121

Slide 121 text

ASD Interventions Research 19 independent studies 27 different interventions

Slide 122

Slide 122 text

No content

Slide 123

Slide 123 text

No content

Slide 124

Slide 124 text

No content

Slide 125

Slide 125 text

No content

Slide 126

Slide 126 text

No content

Slide 127

Slide 127 text

No content

Slide 128

Slide 128 text

No content

Slide 129

Slide 129 text

No content

Slide 130

Slide 130 text

No content

Slide 131

Slide 131 text

No content

Slide 132

Slide 132 text

No content

Slide 133

Slide 133 text

“While everyone is looking at the polls and the storm, Romney’s slipping into the presidency. ”

Slide 134

Slide 134 text

No content

Slide 135

Slide 135 text

Heirarchical modeling

Slide 136

Slide 136 text

Pollster effects

Slide 137

Slide 137 text

No content

Slide 138

Slide 138 text

No content

Slide 139

Slide 139 text

No content

Slide 140

Slide 140 text

No content

Slide 141

Slide 141 text

Data Science

Slide 142

Slide 142 text

Data

Slide 143

Slide 143 text

Science

Slide 144

Slide 144 text

Those who ignore statistics are condemned to re-invent it. -- Brad Efron