Children born during pandemic have lower IQs
"according to the analysis, which is yet to be peer-reviewed."

I'm not sure that's what the study found. Let's do some peer review:
1/

theguardian.com/world/2021/aug… @NatalieGrover As you would have guessed, the study did not measure IQ in the babies. It measured early learning, verbal and non-verbal development.

This is not my field, but call me skeptic that this can be assessed *reliably* in babies <1 year old. What did you do in your first year?

Ik heb de rapporten van het Fieldlab onderzoek doorgenomen. Een verontrustend draadje over bladzijde 11 van "Bijlage 5: Resultaten risico analyse" Of evenementen al dan niet veilig zijn en zo ja onder welke voorwaarden wordt in dit onderzoek bepaald met een risicomodel. Dat is een gebruikelijke methode, maar er zitten haken en ogen aan.

Korte uitleg bij mijn column: waarom is het zo’n probleem wanneer deelnemers zichzelf voor onderzoek kunnen aanmelden?
1/

nrc.nl/nieuws/2021/04… In enquete-onderzoek wil je dat deelnemers een afspiegeling zijn van de *doelpopulatie*, bv alle Nederlanders. *Alleen* als die afspiegeling *representatief* is gelden de resultaten van het onderzoek voor de hele doelpopulatie.

Ik schreef dit weekend een column over slechte wetenschap met daarin een nauwelijks-te-geloven voorbeeld over een index. Hier is een toelichting op de index met wat bronnen.
@nrc @nrcwetenschap
nrc.nl/nieuws/2021/02… Hier is een oud draadje waarin ik de index uitleg en toelicht wat er mis is:

https://twitter.com/cecilejanssens/status/1119256496343269376

Is searching scientific literature the most undervalued aspect of scientific research or is that my impression only?

I gave a lecture to our epi students on how to search literature for their theses.

Here's the essence, incl, I'm biased, how @CoCites makes searching easier.

1/ Everyone who wants to do science needs to find out
- what the state of the art is on their topic, and
- how you set up a study that can move the science forward.
It's not 'any study will do'.

@mauricedehond claimt dat het virus zich voornamelijk verspreidt via inhalering van airborne druppels, waarmee hij lijkt te doelen op aerosols.

De CDC tekst is verwarrend, maar betekent niet wat De Hond denkt.

Hier een overzicht van eerdere versies van de website.

1/

https://twitter.com/mauricedehond/status/1307919320023003137

Het gaat om deze pagina van het CDC: cdc.gov/coronavirus/20…

In de versie van 28 maart was de verspreiding "mainly from person-to-person" en landde de virusdroplet nog op de mond of neus. (bron: Wayback Machine Internet Archive)

@escardio, I'm not a fan of press releases without a link to a conference paper with the methods & results. Your title may be right, but mine may be too.
1/
escardio.org/The-ESC/Press-… The authors talk about naps longer than 60min ... that feels like daytime sleeping to me. They find this predominantly in women and in older people? Is that two subgroup analyses? Could be older women?

Anyway, effects more pronounced in older people rings a bell.

I just learned about second-generation p-values, by @StatEvidence.

They make conceptually more sense as they take into account that the null has an interval too. That there is a difference between null and practically null.
1/
h/t @f2harrell

tandfonline.com/doi/full/10.10… See how they are applied, with the Interval Null in grey.
journals.plos.org/plosone/articl…

Een internist die het allemaal niet zo bijhoudt, maar wel recent gehoord heeft dat er meer beenamputaties waren "terwijl anders alleen hun teen was afgezet". Dat vraagt om een uitleg van de studie (studietje).

Lees even mee.
1/

https://twitter.com/NPORadio1/status/1298307791438974981

De studie is recht-toe-recht-aan: alle vasculaire operaties tussen 16 mrt-30 april in 1 Bredaas ziekenhuis, voor 2020, 2019, en 2018. We lezen wie de patienten waren, welke ziektegeschiedenis en eerdere operaties ze hadden, en welke operatie plaatsvond.
ncbi.nlm.nih.gov/pmc/articles/P…

What I miss in online teaching is the personal connection. So, on Friday I told students that I would stay on Zoom after class for questions.

Several students stayed. For short questions about the course and, to my surprise, for the story behind the gnome in my background.
1/ I told them that the gnome is a gift from my uncle after my partner and I bought our house in the Atlanta. I had joked the house was so cute, it could need a gnome in the garden. And so I got one.
en.m.wikipedia.org/wiki/Garden_gn…

Ik ben niet tegen de corona app, ik ben tegen het invoeren van een maatregel waarvoor bewijs ontbreekt dat het zou kunnen werken. Mijn lat ligt laag: ik hoef geen wetenschappelijke studie, succes in andere landen is genoeg.

1/ thehindu.com/sci-tech/techn… Hier een onderzoekje over de Australische app (te snel ingevoerd, technologisch een ramp).

"The effectiveness of contact tracing apps relies nearly entirely on trust in the government"

innovationaus.com/covidsafe-extr…

Here is a quick summary of my webinar on polygenic risk scores (PRS), hosted by @CDC_Genomics and @MuinJKhoury. A link to all slides and recording is in the last tweet.

The lecture was about PRS, public health, utility, and getting the science right.
1/18 A personal intro: My first work on genetic risk was my MA thesis on why some women perceive their risks of hereditary BC so high that they choose preventive surgery.

This is my family. My mom & grandmom died of BC too young. That's what my aunts wanted to prevent, at high cost.

Terminology in polygenic prediction gets too confusing these days.

What is prediction accuracy? Is it really the correlation with adjusted phenotypes?

And what are confounders?
1/

disq.us/t/3pn7rji Phenotypes were adjusted for age and sex (and others). Are these the confounders? Confounding what?

2/

I read papers on polygenic risk and mental health in the first place as someone who struggles with depression, not as an expert in polygenic risk. Let me share some thoughts about this paper from that perspective.

(Disclosure: I am not having a good day)
1/ Polygenic scores (PRS) studies often rely on drawing inferences about the presence and absence of association based on the the p-value. This is tricky in the UK Biobank as the sample size is large enough to flag tiny effects as statistically significant, as we will see.

CoCites is online!

CoCites is a new way of searching scientific literature. Ideal for (quick) reviews and student theses.
CoCites is embedded in PubMed, as a browser extension for Firefox/Chrome (cocites.com).

People who've tried it, love it.
Here's the essence:
1/ CoCites starts from articles, not keywords, and finds articles on the same specific topic.
2/

In my class on critical thinking, I often discuss topics that I haven't given much thought in my own work.
My attempt to explain *why* p-hacking and cherry-picking are unwise got me confused about how genome-wide association studies work.
1/ When we do hypothesis-driven research, we ideally want to study an association that is sort of in equipoise: the likelihood that the association is present or absent are the same. Then, the results of the study then make the association more or less likely

Some argue that the ROC plot adds no information beyond the AUC. Whether it adds, depends on how much you know about AUC/ROC and on what you want to know. I find them helpful. Here's why:

1. They hint if the sample size is sufficient 2. The differences between curves hint *how* adding predictors improves predictions. If the risk model will be used to identify high-risk indivs, the curves should differ in the lower left corner. Improvements only in the middle of the curve are unlikely relevant.

How do you recognize good research ideas that are worth your time and effort? And how do you evaluate whether the hypotheses of others make sense? Here is my take:
1/ Doing science means following the scientific method. The hypotheses we choose to investigate (also in hypothesis-free research) follows from the rigor of our background research and a process that is called inductive reasoning.

These days, researchers are so focused on data that they seem to forget the value and power of logical reasoning in science.

This week in journal club, we applied logical reasoning to evaluate the strength of hypothesis. Here is a summary.
1/ The most common forms are inductive and deductive reasoning.

In inductive reasoning, we use observations to infer a theory or hypothesis. In deductive reasoning, we use a hypothesis to make predictions about the observations, which are then tested by data.

Last year, I shared a #tweetorial on how to interpret the AUC.
That tweetorial is now published!
With some 'bonus' figures on what the shape of the ROC curve tells about the underlying risk distributions.
1/
academic.oup.com/ije/advance-ar…

https://twitter.com/cecilejanssens/status/1104134423673479169

As explained in the tweetorial: the ROC plot is just a different way of presenting the risk distributions of people with and without the disease of interest

Today in our class on "Critical Thinking: what is the science behind the news?", from @NYTScience :

Does having a dog as a child lower risk of schizophrenia?
1/ I've never seen Cox PH being used this way: case-control data with time-varying exposure, with age at first exposure as timeline. Anyone?

A Kaplan Meier curve for age at being exposed to a dog, with Hazard Ratio.