Guillaume Rousselet

Guillaume Rousselet

@robustgar

Followers3.2K
Following408

open science | data parasite | queue jumper | statistics | visualisation | neuroscience | MEEG | #rstats #matlab #barbarplots EJN & BNA editor

ANA:Governance
Joined on September 05, 2013

Hi there, do you want more Twitter Likes and Followers?

Statistics

We looked inside some of the tweets by @robustgar and here's what we found interesting.

Inside 100 Tweets

Time between tweets:
2 months
Average replies
5
Average retweets
24
Average likes
79
Tweets with photos
33 / 100
Tweets with videos
0 / 100
Tweets with links
0 / 100
To improve the quality of neuroscience & psychology research: let's ban bar graphs from publications.

To improve the quality of neuroscience & psychology research: let's ban bar graphs from publications.

15
89
156

i've got a task where the outcome is a percentage of items correct. items come in 7 batches of 10 items and batches get progressively harder. we stop administering the task if the child can no longer perform it. how might I jointly model "survival" on the task and accuracy?

10
4
9

“most quantities observed in daily life or the sciences cannot take negative values – but all normal distributions assign at least a tiny probability to this impossibility.” https://onlinelibrary.wiley.com/doi/full/10.1111/j.1740-9713.2017.00993.x …

1
3
8

the size of crystals in ice cream follows a lognormal distribution https://academic.oup.com/bioscience/article/51/5/341/243981 … @talyarkoni

3
4
12
JJ
4 days ago

Replying to @robustgar

should be coupled to: Current Incentives for Scientists Lead to Underpowered Studies with Erroneous Conclusions

0
1
1

The natural selection of bad science “Selection for high output leads to poorer methods […] replication slows but does not stop the process of methodological deterioration. Improving the quality of research requires change at the institutional level.” https://royalsocietypublishing.org/doi/full/10.1098/rsos.160384 …

2
11
22
OK this is my new favorite way to present two numbers #dataviz

OK this is my new favorite way to present two numbers #dataviz

28
80
556

Replying to @schotz

@LisaDeBruine and I have materials online from our Master's course in R, Data Skills for Reproducible Science https://gupsych.github.io/data_skills/ 

0
7
15
The log‐normal distribution:
the “Cinderella of distributions”

https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1740-9713.2017.00993.x …

The log‐normal distribution: the “Cinderella of distributions” https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1740-9713.2017.00993.x …

Your regular reminder that *** LIFE IS LOG-NORMAL *** https://stat.ethz.ch/~stahel/lognormal/lnboard/brochure.html …

0
0
4
Folks describing the normal distribution as 'bell-shaped' either have never seen a bell, or they really do not care about the shape of distributions!

Folks describing the normal distribution as 'bell-shaped' either have never seen a bell, or they really do not care about the shape of distributions!

Replying to @robustgar

I had to write it in Spanish for my students.

I had to write it in Spanish for my students.

Yet more lovely null results: no appreciable relationship between strength of hand preference and language ability in 6- to 7-year-old children https://wellcomeopenresearch.org/articles/4-81/v1 …

1
26
57

can’t remind people enough it seems: brain-behaviour correlations cannot be estimated accurately from small sample sizes!

1
32
111

the Society for the Suppression of the Correlation Coefficient http://psychologicalstatistics.blogspot.com/2010/01/when-correlations-go-bad-or-i-always.html …

0
0
4

Replying to @CyrilRPernet @robustgar

Quoted @LorenaABarba

Now out! New report examines Reproducibility and Replicability in Science, with recommendations for researchers, agencies, policy makers, journals, etc. @theNASEM http://ow.ly/BK5350u0Iz2  #ReproducibilityInScience https://twitter.com/LorenaABarba/status/1125740824422174720 …

The new National Academies report adopts these definitions

0
2
4
It’s better not to use reproducibility and replicability interchangeably:

reproducibility = get same results from data and code
replicability = get similar results from new experiment

So most research is not reproducible, let alone replicable.

It’s better not to use reproducibility and replicability interchangeably: reproducibility = get same results from data and code replicability = get similar results from new experiment So most research is not reproducible, let alone replicable.

13
131
340

Quoted @betanalpha

Unpopular opinion: not everyone should do their own statistical analyses, and that should be okay! Forcing unprepared practitioners to develop their own analyses sets them up to fail and reinforces an ecosystem of fragile scientific claims. A thread.

great thread. a big part of the reason I've largely stopped doing data analysis is that I rarely feel sufficiently qualified. it terrifies me to think that most quantitative science claims are made by people with 1 - 2 research methods courses under their belt

12
34
172

I have a dataset where the outcome variable are likelihood ratios, all in the 0-1 range. what’s a distribution for this kind of data?

7
1
6

For those interested in reproducibility (@urfistbordeaux) the #5HTTLPR gene and psychiatric disorders debacle may be a very interesting case study. @robustgar @etienneroesch #ReproducibilityInScience #ScienceIntegrity #Reproducibility

0
4
7
Next Page