Robert Miller

The DFG is a state-funded but private association (e.V.). Its core principle is therefore self-governance - and hardly any of our fellow DFG colleagues seems to be paritculary concerned with revising their infrastructure to improve usability

Albert Ellis who was the first to introduce Cognition into Behavioural Therapy in 1957. (Rational psychotherapy and individual psychology. Journal of Individual psychology, 13) received 707 cites so far according to Google Scholar

.. because it adheres to the closure principle. However I am not completely sure if that generalizes to the intersection hypotheses evaluated by equivalence testing? The additional evaluation of absent super-superiority might result in error inflation?

Question: is there multiplicity when performing equivalence in conjunction with superiority testing? Equivalence is established by the intersection of present non-inferiority and absent super-superiority (term sucks I know). CHMP guidance says switch from non-inferiority to superiority is ok.. (1/2)

Exactly. If more participants are enrolled than needed, a study becomes inherently unethical. If significance and relevance are conceptually collapsible, there is no need for a competition among the different decision frameworks. What's the utility of a "significant" result if it's not relevant?

During the last years, I've started to reframe the problem as a question of relevance that is attributable to significance. A properly planned prospective study ensures that both, significance and relevance coincide. Otherwise any test decision (e.g. for superiority) becomes practically meaningless.

Sorry d = 3

Having taken into consideration, that the variability of the feces ratings will probably be close to 0 (and so will be the correlation), I'd stick to the d = 0.3 ;)

Jingle-jangle detection!

If mean diff is standardized by between-participant variability and the sample is pop-representative, my guess is d = 3 SD

The image presents a bar graph illustrating the smoking rates among adults aged 15 years and older across various countries in Western Europe for the year 2022. The bar heights represent the percentage of adults who used any form of tobacco, excluding e-cigarettes. France has the highest smoking rate at 35%, followed closely by Greece at 33%. Spain follows with 28%, then Belgium at 27%, and Portugal, Switzerland, and Austria, all at 26%. Sweden reports a rate of 23%, while both Italy and Finland have rates of 22%. The Netherlands and Germany each have a smoking rate of 21%. Ireland reports 19%, Denmark at 16%, and the UK and Norway both have the lowest rates at 14%. Additionally, a note indicates that the smoking rates in France and Greece are more than twice as high as those in Denmark, the UK, or Norway. The data source is the World Health Organization, specifically the Global Health Observatory, and the information is labeled as "CC BY," indicating it is licensed for reuse.

Smoking rates vary a lot across Western Europe

At least to my mind.. Anyway - if you like, we might a deep dive into this at the upcoming FGME conf

Indeed. Those serious adverse events (SAE) are extreme manifestations of MDD morbidity. Thus, nocebo treatments (ie waitlists) increase the likelihood of SAE compared to standard of care. Given this risk, a non-inferiority trial against psychotherapy would have been more ethical and informative..

The NEJM article is paywalled, which complicates detailed inspection. Yet, I'd argue that MDD treatments yield potentially fatal adverse events.

Notwithstanding compliance with the WMA Declaration of Helsinki 2000, which requires that participants in therapeutic trials must receive the best proven, non-experimental intervention. Which is certainly not a waitlist..

Wir hatten heute am Institut ein hervorragendes KI-Training für die Psychologie mit vielen praktischen Tipps, durchgeführt von @haesleinhuepf.bsky.social 👌🏼 Die Materialien sind online verfügbar, falls jemand reinschauen möchte: scads.github.io/ai4psycholog...

Terrific @scientificdiscovery.dev post on randomized controlled trials in @ourworldindata.org.

Including this important chart on the impact of pre-registration.

Magically, when people had to pre-register outcomes, many of the benefits disappeared.

ourworldindata.org/randomized-c...

Adverse events in psychotherapy randomized controlled trials: A systematic review Objective: Although evidence for benefits of psychotherapy is substantial, less is known about potential harm. Therefore, we systematically summarized randomized controlled trials (RCTs) to compile...

A #SysReview of adverse events reporting in psychotherapy #RCT -s:
www.tandfonline.com/doi/full/10....

60% of 81 RCTs (since 2007) reported harmful effects.

Based on these data, one can expect adverse events to affect 1/10 (serious adverse events 1/21) patients.

#PsychotherapyResearch

I'd go even further: Psychological placebo conditions, specifically waitlists which serve as nocebos, are inherently unethical (cf WMA Declaration of Helsinki, 2000). Active comparators incl. the simple A/B scheme are the way to go if we are actually into proving the superiority of psychotherapies

The new {marginaleffects} release for #RStats (0.30.0) comes with two new vignettes:

1. Speed up computation with automatic differentiation (often 10x gains) marginaleffects.com/bonus/perfor...

2. Power analyses with {marginaleffects} and {DeclareDesign}. marginaleffects.com/bonus/power....

Models as Prediction Machines: How to Convert Confusing Coefficients into Clear Quantities Abstract Psychological researchers usually make sense of regression models by interpreting coefficient estimates directly. This works well enough for simple linear models, but is more challenging for more complex models with, for example, categorical variables, interactions, non-linearities, and hierarchical structures. Here, we introduce an alternative approach to making sense of statistical models. The central idea is to abstract away from the mechanics of estimation, and to treat models as “counterfactual prediction machines,” which are subsequently queried to estimate quantities and conduct tests that matter substantively. This workflow is model-agnostic; it can be applied in a consistent fashion to draw causal or descriptive inference from a wide range of models. We illustrate how to implement this workflow with the marginaleffects package, which supports over 100 different classes of models in R and Python, and present two worked examples. These examples show how the workflow can be applied across designs (e.g., observational study, randomized experiment) to answer different research questions (e.g., associations, causal effects, effect heterogeneity) while facing various challenges (e.g., controlling for confounders in a flexible manner, modelling ordinal outcomes, and interpreting non-linear models).

Figure illustrating model predictions. On the X-axis the predictor, annual gross income in Euro. On the Y-axis the outcome, predicted life satisfaction. A solid line marks the curve of predictions on which individual data points are marked as model-implied outcomes at incomes of interest. Comparing two such predictions gives us a comparison. We can also fit a tangent to the line of predictions, which illustrates the slope at any given point of the curve.

A figure illustrating various ways to include age as a predictor in a model. On the x-axis age (predictor), on the y-axis the outcome (model-implied importance of friends, including confidence intervals). Illustrated are 1. age as a categorical predictor, resultings in the predictions bouncing around a lot with wide confidence intervals 2. age as a linear predictor, which forces a straight line through the data points that has a very tight confidence band and 3. age splines, which lies somewhere in between as it smoothly follows the data but has more uncertainty than the straight line.

Ever stared at a table of regression coefficients & wondered what you're doing with your life?

Very excited to share this gentle introduction to another way of making sense of statistical models (w @vincentab.bsky.social)
Preprint: doi.org/10.31234/osf...
Website: j-rohrer.github.io/marginal-psy...

In an experiment, Pew Research Center demonstrated that opt-in and probability-based surveys produced very different results about young adults' views of the Holocaust and abortion.

Remember, if you encounter what seems like an implausible survey finding, ask:
1. Were survey respondents selected randomly or was this an opt-in poll?
2. Could the results, especially for young adults, be driven by bogus respondents?

Keep this post in mind: www.pewresearch.org/short-reads/... 🧪

GAMLSS, NHANES, and my own personal hell – Notes from a data witch Fiiiiiiinally she writes the cursed GAMLSS post. Oh and it’s also about the NHANES data set I guess

It’s been a little overshadowed by the slightly unexpected mental health issues that hit me very hard yesterday afternoon, but for what it’s worth here’s the GAMLSS regression post I’ve been working on for the past month #rstats

blog.djnavarro.net/posts/2025-0...

A Survey of 1144 ECT Recipients, Family Members and Friends: Does ECT Work? The last placebo-controlled ECT trial for depression occurred in 1985. While awaiting trials that meet today's standards of evidence-based medicine, this paper presents the responses, to an online su...

"It is striking that nearly half (49%) reported that their quality of life was made ‘much worse’ (22%) or ‘very much worse’ (27%) by ECT." 🤔

For those interested, here is a link to a new power paper:

Hancock, G. R., & Feng, Y. (2026). nmax and the quest to
restore caution, integrity, and practicality to the sample size planning process. Psychological Methods.

yifengquant.github.io/Publications...

Assurance in clinical trial design Conventional clinical trial design involves considerations of power, and sample size is typically chosen to achieve a desired power conditional on a specified treatment effect. In practice, there is ...

​ onlinelibrary.wiley.com/doi/abs/10.1... EDR and assurance become exchangeable when the power of a study just equals the average power of all meta-analyzed past studies and when integrating across the entire posterior distribution of the effect size.

A DuckDB-Backed Version of dplyr A drop-in replacement for dplyr, powered by DuckDB for performance. Offers convenient utilities for working with in-memory and larger-than-memory data while retaining full dplyr compatibility.

And if you're not particularly enthusiastic about writing your own SQL, there's this:

How about using the term "assurance" to this end, as proposed in medical biometry?

New blog post! Why experiments are the gold standard for answering causal questions (pedermisager.org/blog/why-exp...). Many text books insist on experimental evidence to draw causal inferences bvut don't fully explain exactly what gives experiments their special powers.