Daily data science tweets from @JohnDCook.
Houston
- Tweets 11,458
- Following 19
- Followers 196,000
- Likes 122
Pinned Tweet
“Statistics is hard, especially when effects are small and variable and measurements are noisy. There are no quick fixes … and a formulaic approach to statistics is a principal cause of the current replication crisis.” — Blakeley McShane
3
7
27
4,025
Put Santa hats on estimated quantities for Christmas.
\usepackage{realhats}
...
\hat[santa]{y} = \beta_0 + \beta_1 x
6
104
801
50,899
“There is no point in being precise when you don’t know what you’re talking about.” — John Tukey
8
165
659
'Errors using inadequate data are much less than those using no data at all.' -- Charles Babbage
21
390
672
“When you’re fundraising, it’s AI.
When you’re hiring, it’s ML.
When you’re implementing, it’s logistic regression.”
11
177
673
"Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing 'look over there'." -- xkcd
8
102
587
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
15
86
585
Instead of calling some data "real world data", we could just call that "data," and call everything else "artificial world data."
28
114
568
Bayes’ theorem implies
0 prior probability + any data = 0 posterior probability
No data can change your mind unless you think there’s a small chance that you might be wrong.
#CromwellsRule
6
88
520
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
5
98
477
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
11
50
520
45,862
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
8
44
471
43,862
The number of people who can carry out statistical calculations has grown faster than the number of people who can interpret the results.
4
114
439
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
5
99
393
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
17
45
388
87,575
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
13
73
374
“Mathematicians, pure and applied, think there is something weirdly different about statistics. They are right.” -- James Franklin
5
91
366
Sample variance is an unbiased estimator of variance, but sample standard deviation is a biased estimator of standard deviation.
3
35
343
45,527
Good statistical practice can increase your chances of being correct while decreasing your chances of being published. Conflict of interest.
3
58
352
“Mathematicians, pure and applied, think there is something weirdly different about statistics. They are right.” -- James Franklin
10
62
331
Statistical Modeling: Regression, Survival Analysis, and Time Series Analysis
Free textbook, 676 page PDF
math.wm.edu/~leemis/otext.pd…
1
113
352
73,621
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
9
50
324
A BAYESIAN CREDIBLE INTERVAL IS A PROBABILITY STATEMENT ABOUT THE PARAMETER THETA WHEREAS A FREQUENTIST CONFIDENCE INTERVAL IS NOT BUT RATHER A STATEMENT ABOUT THE PROCESS USED TO CREATE THE INTERVAL AND BESIDES FREQUENTIST STATISTICS CANNOT MAKE PROBABILITY STATEMENTS ABOUT PAR
4
57
318
'A Bayesian is one who, vaguely expecting a horse, and catching a glimpse of a donkey, strongly believes he has seen a mule.'
4
233
293
'Statistics is the art of stating in precise terms that which one does not know.' -- William Kruskal
2
58
303
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
2
85
304
Data scientists often complain that the bulk of their work is data cleaning.
But if you see data cleaning as the work, not just an obstacle to the work, it can be interesting.
You could think of it as data pathology, a kind of analysis before the intended analysis.
7
62
291
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
1
66
288
A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction models
sciencedirect.com/science/ar…
11
129
271
“He uses statistics as a drunken man uses lampposts, for support rather than illumination.” — Andrew Lang
1
65
271
Type I error: False positive
Type II error: False negative
Type III error: Answering the wrong question
6
181
271
'Bias is a learner's tendency to consistently learn the same wrong thing.' -- Pedro Domingos
4
76
281
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
7
80
273
"For every complex problem, there is a solution that is simple, neat, and wrong." -- H. L. Mencken
1
48
256
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
9
86
262
Things every child needs to hear.
!. I love you.
2. I'm proud of you.
3. A p-value is NOT the probability that the alternative hypothesis is false.
2
42
277
15,181
A small p-value means the data you saw were unlikely, assuming the null hypothesis AND all your implicit assumptions.
4
42
257
Good statistical practice can increase your chances of being correct while decreasing your chances of being published. Conflict of interest.
5
126
245
“The current boom in artificial intelligence is primarily driven by just two equations – gradient descent and logistic regression – put together to create what is known as a neural network.” — David Sumpter
5
57
250
"One does not judge a model on whether it is wrong, rather how costly the error." -- @nntaleb
1
110
250
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
11
36
247
58,122
'It is easy to lie with statistics; it is easier to lie without them.' -- Frederick Mosteller
3
65
241
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
5
46
227
22,774
"Every researcher using statistics insists correlation isn’t causation, until it comes to HIS data."-- Matt Briggs
4
39
238
'The year 2000 was essentially the point at which it became cheaper to collect information than to understand it.' -- Freeman Dyson
1
62
233
Good statistical practice can increase your chances of being correct while decreasing your chances of being published. Conflict of interest.
80
226
You always have prior information before you do an experiment, because *something* motivated you to do the experiment.
4
39
219
'Statistics is the art of stating in precise terms that which one does not know.' -- William Kruskal
1
63
223
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
2
63
218
'Statisticians, like artists, have the bad habit of falling in love with their models.' -- George Box
81
219
Naive Bayes is called 'naive' because it assumes things are independent that definitely are not. And in the right setting works well anyway.
3
33
210
'It is easy to lie with statistics; it is easier to lie without them.' -- Frederick Mosteller
2
53
204
'A Bayesian is one who, vaguely expecting a horse, and catching a glimpse of a donkey, strongly believes he has seen a mule.'
8
114
194
You always have prior information before you do an experiment, because *something* motivated you to do the experiment.
15
50
12,153
"I've reproduced 130+ research papers about predicting the stock market, coded them from scratch and recorded the results. Here's what I've learnt."
teddit.net/r/algotrading/com…
5
65
200
"People bred, selected, and compensated to find complicated solutions do not have an incentive to implement simplified ones." -- @nntaleb
5
105
187
One way to think of the Markov assumption: The future is independent of the past, given the present.
4
43
189
'Statistics is the art of stating in precise terms that which one does not know.' -- William Kruskal
1
41
192
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
6
117
193
Lawrence Leemis’ textbook Statistical Modeling contains a version of his famous distribution diagram specialized for lifetime distributions.
math.wm.edu/~leemis/otext.pd… [pdf]
1
46
193
32,268
“There is no point in being precise if you do not even know what you are talking about.” –John von Neumann
2
50
191
'Statistics is the art of stating in precise terms that which one does not know.' -- William Kruskal
1
28
199
Good statistical practice can increase your chances of being correct while decreasing your chances of being published. Conflict of interest.
6
49
193
One way to think of the Markov assumption: The future is independent of the past, given the present.
4
31
191
No amount of data can persuade you of something you assign zero prior probability to. -- Cromwell's rule
9
54
183
A p-value is NOT the probability that the null hypothesis is correct. Not even close. Persistent and pernicious misunderstanding.
8
58
183
Most medieval castles were made of wood. We think most were made of stone because of survivor bias.
16
436
180