Stephen John Senn @stephensenn Twitter profile

Pinned Tweet

Stephen John Senn

3 years

Your inability to think statistically is what makes you lurch from unjustified premise to illogical conclusion via irrelevant data.

13

46

224

Last Seen Profiles

@jandakembangstw

@RommelHern93479

@delfiiappendino

@meanwhilelife

@jandakembangstw

@Tootta69Q

@noisemusicstan

@sabriam_

@ripalekk

@Tootta69Q

@justaboutshows

@indiainbrazil

@katttesquad

@crxmsenVal

@thepdfc

@MwydHazmy38935

@i_sss_ai

@dokudoku884

@ezgipeksen

@beetalqa9eed

@couples_mot7arr

@Tootta69Q

@Doi_Washoku

@iidabii_

@CwmWanderersAFC

@mizuna_life

@Uweueuexoxo8714

@Susy_sintabu

@FiansoGonzalez

@oJaehyunct

@minaszns

@koudai_unmask

@al91la

@MuhammdRudy_

@perotoss400

Stephen John Senn

@stephensenn

10 years

I've been studying statistics for over 40 years & I still don't understand it. The ease with which non-statisticians master it is staggering

32

454

1K

Stephen John Senn

@stephensenn

4 years

I am a medical statistician who occasionally dabbled in infectious diseases but I know that I’m not remotely near being an expert on the pandemic. Luckily almost everybody else on Twitter is.

20

72

716

Stephen John Senn

@stephensenn

5 years

@annaeveryday Flight attendant: is there a doctor on board? Statistician: did you ask us before we took off?

3

40

555

Stephen John Senn

@stephensenn

5 years

If more scientists calculated confidence distributions, they would realise that a point estimate is a value in which you have zero confidence.

21

103

368

Stephen John Senn

@stephensenn

4 years

@neilpearce53x Data science is the art of producing misleading conclusions from huge datasets whereas statistics is the much more efficient science of doing so from small ones.

8

114

334

Stephen John Senn

@stephensenn

6 years

My latest piece on #personalizedmedicine , Statistical pitfalls of personalized medicine cites @VPplenarysesh

Statistical pitfalls of personalized medicine

Nature - Misleading terminology and arbitrary divisions stymie drug trials and can give false hope about the potential of tailoring drugs to individuals, warns Stephen Senn.

www.nature.com

7

156

310

Stephen John Senn

@stephensenn

3 years

It’s here @Wiley_Stats

10

27

291

Stephen John Senn

@stephensenn

4 years

Just because you are modelling coronavirus does not mean you can’t get it. There is no nerd immunity.

5

54

271

Stephen John Senn

@stephensenn

3 years

You can be rich in data but poor in information.

7

40

257

Stephen John Senn

@stephensenn

4 years

Personally, I think I did my bit for social distancing years ago when I became a statistician.

7

27

249

Stephen John Senn

@stephensenn

2 years

"RCTs neither guarantee nor require balance. Imbalance is inevitable ... randomisation is a device for enabling us to make precise estimates of an inevitable imprecision."

S. Senn: Randomisation is not about balance, nor about homogeneity but about randomness (Guest Post)

Stephen Senn Consultant Statistician Edinburgh The intellectual illness of clinical drug evaluation that I have discussed here can be cured, and it will be cured when we restore intellectual primac…

errorstatistics.com

6

58

237

Stephen John Senn

@stephensenn

4 years

@daniela_witten When working for pharma I was told by a medical investigator that I was analysing cross-over trials all wrong and needed to contact an academic. I gave them a copy of The look on their face when they realised it was by me was priceless.

7

8

224

Stephen John Senn

@stephensenn

3 years

Just realised that the amazing living legend of statistics CR Rao turned 101 a few days ago. I saw him receive the Guy medal in Gold in 2011. The paper he wrote 25 years ago is of great interest to any student of the history of statistics.

3

33

221

Stephen John Senn

@stephensenn

3 years

Repeat after me. Randomisation is not meant to guarantee balance. Nothing can as Fisher explains here. Randomisation permits valid estimation of the possible effect of imbalance.

10

51

217

Stephen John Senn

@stephensenn

3 years

Just received this by ⁦ @rlmcelreath ⁩ and am looking forward to learning from it.

8

15

212

Stephen John Senn

@stephensenn

2 years

“The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.” John Tukey,

7

38

210

Stephen John Senn

@stephensenn

4 years

You can divide data-analysts into two. Those who think ML stands for machine learning and those who know it means maximum likelihood.

7

35

200

Stephen John Senn

@stephensenn

3 years

What RA Fisher did not say 1) "Randomisation balances all confounding factors" What RA Fisher did say 2) "Randomisation permits valid estimation of the uncertainty due to unknown confounders" What SJS says. "If you maintain 1 is the point, keep reading until you understand 2."

4

36

192

Stephen John Senn

@stephensenn

6 years

We need to teach statistics as a form of advanced common sense. What happens is that users kick their brains out of reasoning mode to magical recipe mode as soon as they start stats.

Jack Wilkinson

@jd_wilko

6 years

When you do a lot of statistical consultancy, it is very difficult not to come to the conclusion that most people doing medical research do not understand the basics.

11

19

114

5

66

190

Stephen John Senn

@stephensenn

4 years

A key difference between mathematics and statistics is that whereas the former is full of lemmas the latter is full of dilemmas

4

34

183

Stephen John Senn

@stephensenn

2 years

If I wrote a book about regression modelling, I would be tempted to call it The Book of Y.

16

10

177

Stephen John Senn

@stephensenn

6 years

A reminder to all biostatisticians: we are the bad fairies of medical research: nobody really likes us, they don’t want us at the party, when we stomp in we are disagreeable and we finish by sending everyone to sleep.

15

33

172

Stephen John Senn

@stephensenn

3 years

A medical statistician is one who classifies mankind into two groups: "dead" and "not yet dead".

9

24

170

Stephen John Senn

@stephensenn

5 years

@ColinJCarlson When you have grown sufficiently in statistics it is explained to you that the distinction between fixed and random effects is a) absolutely crucial and b) somewhat arbitrary.

3

19

161

Stephen John Senn

@stephensenn

4 years

Personalised medicine has been making the same stunning contribution to controlling the pandemic it makes to healthcare generally.

4

21

168

Stephen John Senn

@stephensenn

2 years

This is my view on P-values and multiplicity

11

35

156

Stephen John Senn

@stephensenn

4 years

It is the duty (and pleasure) of every Swiss to eat several kg of chocolate, The Matterhorn is one of our three iconic mountains, but I reached the age of 67 before I noticed the bear in this picture.

16

154

Stephen John Senn

@stephensenn

2 years

1/2) Tips for those who wish to write about RCTs 1) Few trials are completely randomised 2) True binary outcomes are rare 3) Random sampling does not occur 4) Study effects can be important 5) Estimation of uncertainty is vital 6) Patients do not arrive simultaneously

1

33

150

Stephen John Senn

@stephensenn

3 years

Just arrived

2

20

152

Stephen John Senn

@stephensenn

11 years

Your inability to think statistically is what makes you lurch from unjustified premise to ilogical conclusion via irrelevant data.

7

52

147

Stephen John Senn

@stephensenn

4 years

The bootstrap: A means to compensate for the possible unrepresentative nature of your sample by pretending it’s the population.

4

19

140

Stephen John Senn

@stephensenn

5 years

Precision medicine: a field that aims to replace the tired simplistic "one-size fits all" philosophy with the exciting new "no size fits anybody" philosophy.

5

31

143

Stephen John Senn

@stephensenn

3 years

There seems to be an outbreak of mass wisteria in Edinburgh.

4

8

142

Stephen John Senn

@stephensenn

6 months

Tutorial paper on n-of-1 trials now available.

The analysis of continuous data from n-of-1 trials using paired cycles: a simple tutorial - Trials

N-of-1 trials are defined and the popular paired cycle design is introduced, together with an explanation as to how suitable sequences may be constructed.Various approaches to analysing such trials...

trialsjournal.biomedcentral.com

4

39

142

Stephen John Senn

@stephensenn

3 years

Seven myths of randomisation. Even statisticians fall for number 7.

6

39

135

Stephen John Senn

@stephensenn

3 years

1/2) My pet hates in statistical design & analysis (in no particular order) 1) Analysing ordered categorical data as if unordered 2) Arbitrary dichotomies 3) Using change from baseline rather than ANCOVA 4) Responder analysis 5) Minimisation 6) Propensity scores

12

27

134

Stephen John Senn

@stephensenn

5 years

Warning to all trialists. If you tell me the treatment failed because the placebo response was too high I shall file you under “idiot”.

8

25

129

Stephen John Senn

@stephensenn

2 years

The hidden effect of time This gem by Doug Altman & Patrick Royston should be read by all who think that clinical trials are representative and who have difficulty understanding concurrent control.

The hidden effect of time

It is customary to regard datasets as homogeneous with respect to the order of collection of the measurements. Examples are given in which this assumption is breached. Hidden time trends have...

onlinelibrary.wiley.com

5

30

129

Stephen John Senn

@stephensenn

9 months

This is probably true.

11

22

128

Stephen John Senn

@stephensenn

7 months

I’m baffled & exasperated by the numbers doing statistical analysis who imagine that the patients in a clinical trial can be regarded as a random sample from the population defined by the inclusion criteria. Some heinous crime was committed in educating them.

15

22

130

Stephen John Senn

@stephensenn

2 years

On meta-analysis "It is not just a question of pushing data into some software sausage machine and waiting for a summary to appear. Empowering the statistically innocent to perform statistical analyses has its drawbacks."

Overstating the evidence – double counting in meta-analysis and related problems

The problem of missing studies in meta-analysis has received much attention. Less attention has been paid to the more serious problem of double counting of evidence.Various problems in overstating...

www.ncbi.nlm.nih.gov

4

28

130

Stephen John Senn

@stephensenn

2 years

As an act of statistical homage, of course.

8

3

121

Stephen John Senn

@stephensenn

2 years

@Theresa_Chapple @MaartenvSmeden I didn’t. I decided to be a scientific parasite (aka as a statistician). That’s much easier. Instead of having to collect data I get to insult others for what they do with them.

6

5

120

Stephen John Senn

@stephensenn

5 years

A few weeks back @f2harrell kindly suggested that I should put up an index to my blogs. Here is a first attempt Many of these are hosted by @learnfromerror , to whom I am also very grateful #Statistics #clinicaltrials

3

37

119

Stephen John Senn

@stephensenn

6 years

I believe it was Leonard Cohen who wryly remarked that in his youth he suffered from the delusion that folk songs defeated the Nazis. I am reminded of this whenever I am told about big data defeating cancer.

4

34

117

Stephen John Senn

@stephensenn

6 years

Look: no algebra. Regression to the mean in one slide. @StatsJen

2

31

115

Stephen John Senn

@stephensenn

7 years

My latest blog on randomisation . My thanks to @learnfromerror for hosting it and to @trishgreenhalgh for alerting me to the paper by Deaton and Cartwright

S. Senn: Being a statistician means never having to say you are certain (Guest Post)

Stephen Senn Head of Competence Center for Methodology and Statistics (CCMS) Luxembourg Institute of Health Twitter @stephensenn Being a statistician means never having to say you are certain A re…

errorstatistics.com

0

63

115

Stephen John Senn

@stephensenn

4 years

Unpopular point of view. Beamer is just PowerPoint for nerds and does not make presentations better.

22

7

115

Stephen John Senn

@stephensenn

8 years

If you think P-values exaggerate significance, have you tried university press-offices?

3

40

115

Stephen John Senn

@stephensenn

2 years

@ancetetere @statsepi Keep fighting the good fight.

Statistical pitfalls of personalized medicine

Nature - Misleading terminology and arbitrary divisions stymie drug trials and can give false hope about the potential of tailoring drugs to individuals, warns Stephen Senn.

www.nature.com

2

21

113

Stephen John Senn

@stephensenn

2 years

I don't eat steaks rare, seafood raw or do weights but I worry about type I errors. That's why I am an alpha male.

10

14

114

Stephen John Senn

@stephensenn

4 years

When I stop to consider, statistics is a subject where you get to play with other people’s data while insulting their intelligence. It’s baffling why I enjoy it so much.

4

15

109

Stephen John Senn

@stephensenn

6 years

A statistician I knew once replied to a brain surgeon whose hobby was statistics that hers was brain surgery.

David Robinson

@drob

6 years

Saying "no one should analyze data without consulting a statistician, they could misinterpret the results" is like saying "no one should exercise without a personal trainer, they could injure themselves."

25

87

485

7

38

111

Stephen John Senn

@stephensenn

4 years

Of course, the burning question to which the general public really wants the answer is “ are these corona virus🦠 models frequentist or Bayesian?”

4

20

108

Stephen John Senn

@stephensenn

6 years

I have loathed and detested stepwise regression ever since I first encountered it 40 years ago.

Edward Tufte

@EdwardTufte

6 years

If use of every statistical model was accompanied warnings similar to those accompanying every prescription drug. In medical research, mistakes in data analysis can shorten 1000s of lives. #biostatistics #statistics #datascience

11

468

913

2

45

108

Stephen John Senn

@stephensenn

5 years

Here’s a terrible truth about statistics: not only is correlation not always causation, sometimes it isn’t even correlation.

Stephen Senn: Rothamsted Statistics meets Lord’s Paradox (Guest Post)

Stephen Senn Consultant Statistician Edinburgh The Rothamsted School I never worked at Rothamsted but during the eight years I was at University College London (1995-2003) I frequently shared a tra…

errorstatistics.com

1

44

107

Stephen John Senn

@stephensenn

4 months

This is your regular reminder that the analysis of randomised experiments does not require all unknown confounders to be balanced. It make an allowance for the fact that they won't be. If we knew they were all balanced, the standard analysis would be wrong.

4

24

111

Stephen John Senn

@stephensenn

2 years

I have been developing a four part clinical trials course. The first part has also been written up as a self-instruction document. It is available here . I hope that the link works. #clinicaltrials #statistics

7

21

107

Stephen John Senn

@stephensenn

5 years

Couldn’t we just rebrand Bayesianism as personalised statistics?

8

11

108

Stephen John Senn

@stephensenn

4 years

There are two traps to avoid in acting as a consultant statistician: Trap 1 Giving your clients what you want Trap 2 Giving your clients what they want

5

26

107

Stephen John Senn

@stephensenn

4 years

I have a risqué joke about survival analysis but it’s been censored

5

106

Stephen John Senn

@stephensenn

3 years

Statistical Issues in Drug Development 3rd edition. Coming shortly, I hope (Was promised much earlier: the delays are my fault.)

6

13

104

Stephen John Senn

@stephensenn

5 years

1) If covariates are imbalanced but not prognostic, the inference from RCTs will not change by conditioning 2) If covariates are balanced but prognostic, the inference will change by conditioning 3) If you babble on about 'balance' in RCTs you must sit on the stupid step

4

26

104

Stephen John Senn

@stephensenn

6 years

Free stats lesson. See these four “quarters” here. The three spaces between them are the quartiles.

3

38

104

Stephen John Senn

@stephensenn

3 years

"On average, when statisticians say 'mean', they mean average." Sayings of Confuseus

10

25

100

Stephen John Senn

@stephensenn

4 years

Given that Pfizer said they would be using a Bayesian analysis, what the whole world is waiting for is not to get the results but what the prior distribution is.

8

13

100

Stephen John Senn

@stephensenn

4 years

If all the very many people who claim to dislike and mistrust statistics would also agree not to use them, life would be much simpler.

5

24

99

Stephen John Senn

@stephensenn

4 years

To criticise randomisation because it does not guarantee balance is like criticising a bridge-builder because thermal expansion means the bridge will stretch, while overlooking that the design explicitly allows for this fact.

4

14

100

Stephen John Senn

@stephensenn

3 years

My attempt at producing a video. Being wrong about randomisation v2 via @YouTube duration about 6 minutes (Plot spoiler: randomisation is not about balance.)

Being wrong about randomisation v2

A short video on randomisation

www.youtube.com

6

26

99

Stephen John Senn

@stephensenn

2 years

One of the most depressing phrases in analysis of clinical trials: “subgroup analysis”.

15

11

99

Stephen John Senn

@stephensenn

7 years

I am very very bored with people blaming the so-called replication crisis on P-values

10

25

97

Stephen John Senn

@stephensenn

6 years

So long and thanks for all the Poisson regression (don't forget the over-dispersion) #retiredbutnotshy

7

1

94

Stephen John Senn

@stephensenn

1 month

An elderly statistician rambling on about crossover trials.

8

2

95

Stephen John Senn

@stephensenn

2 years

Just arrived! This looks very interesting. My thanks to @AndrewPGrieve

1

18

91

Stephen John Senn

@stephensenn

3 years

No comment.

2

19

89

Stephen John Senn

@stephensenn

4 years

1/2) I am thinking of writing a blog with the title ‘Actually, randomisation is not about balance it’s about randomness’

9

13

91

Stephen John Senn

@stephensenn

4 years

The first rule of multivariable modelling club is you do not call it multivariate modelling club.

1

8

91

Stephen John Senn

@stephensenn

4 years

@kareem_carr I was initially intrigued by this question but now I can’t be bothered to probit

1

0

92

Stephen John Senn

@stephensenn

6 years

2019 is going to be one hell of a year in clinical research "By the year 2020, gene-based designer drugs are likely to be available for conditions like diabetes, Alzheimer's disease, hypertension, and many other disorders."

12

34

88

Stephen John Senn

@stephensenn

6 years

Bayesian books, eh? If I read a few do I get a good idea of what’s in the rest?

7

89

Stephen John Senn

@stephensenn

11 months

In your other news, computer scientists have invented a new technique for optimising clinical trial analysis called covariate adjustment.

7

11

87

Stephen John Senn

@stephensenn

6 years

Industry is more alarmed about reproducibility than academia

7

41

88

Stephen John Senn

@stephensenn

6 years

Client: I know that you should consult the statistician before you collect the data but ... Statistician: no problem. Let's arrange to meet last year.

0

14

87

Stephen John Senn

@stephensenn

6 years

Four blogs on randomisation by me 1) 2) 3) 4) thanks to @learnfromerror for making them possible.

S. Senn: Fishing for fakes with Fisher (Guest Post)

Stephen Senn Head of Competence Center for Methodology and Statistics (CCMS) Luxembourg Institute of Health Twitter @stephensenn Fishing for fakes with Fisher Stephen Senn The essential fa…

errorstatistics.com

1

24

86

Stephen John Senn

@stephensenn

11 months

This has arrived and looks very good.

3

9

84

Stephen John Senn

@stephensenn

2 years

Pina Collider, presumably.

2

8

86

Stephen John Senn

@stephensenn

4 years

My back of the envelope calculations on the Moderna trial

6

30

84

Stephen John Senn

@stephensenn

3 years

I was getting gloomy about climate change but then I was told that big data and machine learning will be leveraged through artificial intelligence to deal with it and now the sun is shining (albeit rather hotly).

7

10

80

Stephen John Senn

@stephensenn

4 years

TBF, it's an easy mistake for a Government minister to assume that 'World-beating' means Excel.

2

13

81

Stephen John Senn

@stephensenn

4 years

@GeorgeGreenwood Because they ran out of parsnips?

3

0

83

Stephen John Senn

@stephensenn

4 years

If you think that Normality is a more important assumption that independence, you need to go back to stats school. And if you think that doing a rank test will deal with a lack of independence you must sit on the naughty step.

5

10

83

Stephen John Senn

@stephensenn

3 years

The Design of Experiments section. Fisher’s contribution is absent because it lives on the Fisher shelf.

12

9

83

Stephen John Senn

@stephensenn

4 years

If you like using propensity matching, think again. See or by @PhilDawid or by @kinggary

Why Propensity Scores Should Not Be Used for Matching

Gary King and Richard Nielsen. 2019. “Why Propensity Scores Should Not Be Used for Matching.” Political Analysis, 27, 4, Pp. 435-454.

gking.harvard.edu

3

25

79

Stephen John Senn

@stephensenn

3 months

Time to quote Confuseus again. “ Enjoy the result you have found by exploratory analysis, for you will not find it again.”

5

14

81

Stephen John Senn

@stephensenn

2 years

“Physicians suffer from dichotomania whereas statisticians have continuitis.” Sayings of Confuseus

5

13

79

Stephen John Senn

@stephensenn

6 years

Statistician. I ordered IPA and you’ve given me a bitter, you’ll have to change it. Barman. Aren’t you a statistician? Statistician. Yes. Barman. You are supposed to proceed as if the intended assignment was actually delivered. #intentiontotreat

0

25

81

Stephen John Senn

@stephensenn

3 years

Perfect timing. I look forward to reading this by ⁦ @d_spiegel ⁩ and ⁦ @anthonybmasters ⁩