Dear reader,

This post is a quick, practical update: I’m heading back to Europe, and this time the “why” is statistics. Not vibes, not a vague “growth journey.” Actual methods. Actual data. Actual standard errors that refuse to be ignored.

What happened last time

Last time I was in Turkey, I came home with the same feeling I always get after being around smart people doing real work: I saw how much I still had to learn, and I liked that. A lot.

Europe has a deep bench when it comes to measurement, survey design, and large-scale assessment work, the kind of work where the data are messy on purpose because the real world is messy on purpose. If you care about education data (PISA, TIMSS, PIRLS, national exams, program evaluation), you end up caring about things like complex sampling, weighting, missingness, and validity evidence. So… here we are.

Why Europe

Part of the appeal is that the applied problems are honest. You’re not just fitting models to tidy spreadsheets. You’re dealing with multi-stage samples, cluster effects, replicate weights, and constructs that do not behave like obedient little Gaussian puppies.

I’m also excited about the ecosystem: methods workshops, summer schools, and research groups where psychometrics and “classic” statistics actually talk to each other. That’s where a lot of the good stuff happens, especially around measurement invariance, fairness, and how we summarize ability or achievement when we only observe noisy proxies.

What I’m working on

My main focus is learning and shipping. Concretely:

• Survey inference: how weights, stratification, and clustering change estimation and uncertainty, and how to do it correctly without turning everything into a black box.
• Plausible values: what they are, why they exist, how to combine them, and how to keep your interpretation sane when you’re explaining results to humans.
• Psychometric modeling: IRT, latent variable models, and the practical “so what” side of reliability and validity when the stakes are real.
• Reproducible workflows: clean pipelines, versioned outputs, and code that future-me will not curse at 2 a.m.

Some of this will turn into blog posts. Some will become small tutorials. Some will just be me quietly building competence, which is honestly underrated.

How I’m approaching it

I’m trying to keep a simple rule: don’t confuse cleverness with correctness. If a method is “cool” but I can’t explain its assumptions, failure modes, and interpretation, then it’s not ready to touch real decisions.

So I’m going in with a notebook mindset: read, implement, test on toy data, test on real data, write down what breaks, repeat. It’s not glamorous, but it works.

Happy estimating,

- Fatih

P.S. If you want to chat about survey stats, psychometrics, or anything measurement-related, feel free to reach out via email.