Mirko

#academicTwitter is dead, and I'm tired of being constantly fed Musk's delirious tweets and interviews. I'm leaving X for the near future, or maybe for good. If you want to contact me, do it on #LinkedIn or by email. #byeBye

In most places in the world, power from new renewables is now cheaper than power from new fossil fuels. Why did renewables become so cheap so fast? The answer: learning curves. For renewables and other technologies that follow learning curves, with each doubling of the cumulative installed capacity their price declines by the same fraction. The price of electricity from fossil fuels, however, does not follow learning curves. This means we should expect that the price difference between expensive fossil fuels and cheap renewables will become even larger in the future.

📢 I am hiring! 📢 I have a vacancy for a #postdoc in #statistics (longitudinal data analysis) in our group at #Leiden University. Interested? Here you can find more information on the position and selection criteria: universiteitleiden.nl/vacatures/2024…

@Chicca57105111 @lobliquo @ultimora_pol Veramente no, un candidato può perdere le presidenziali anche se ha preso più voti. Vedi Clinton Vs Trump nel 2016

#USA2024 - Risultati finali: #Trump vince in tutti gli stati in bilico. A livello nazionale ha ottenuto il 50,5% dei voti contro il 47,9% di Harris. (Foto BNO) @ultimora_pol

I'm happy to share that the talk about the #R package #pencal that I presented at @_useRconf in July is now available on Youtube: youtu.be/jnpiSldbTko #Rstats #statistics #machineLearning #software #prediction #survivalhorror

@EU_Eurostat 1.1 billion

Let's mark European Statistics Day with a COMPETITION 🥳📊 How many individual data points are there in Eurostat’s online database❓ Write your guess below and the 3 first closest guesses win Eurostat prizes. 🏆 Competition closes on 21/10/24 at 11:00 CEST. #EurStatsDay

I'm on my way to Bochum for #bsimsworldcongress. Tomorrow at 14.25 I will talk about a few methods for the dynamic prediction of survival outcomes when you have numerous longitudinal predictors: #paperID389" target="_blank" rel="nofollow noopener">conftool.com/bernoulli-ims-…

@EikoFried A #stats reference on this is: projecteuclid.org/journals/stati…

@EikoFried A prediction model doesn't have to be an explanatory model though. Rather than aiming to identify the process that may generate the data, it aims to maximize predicive performance 1/2

1/2 I'm looking for literature on the importance of theory building and testing for black box machine learning prediction. In psych, these are discussed as opposites, but it's obvious that *choosing* a black box model relies on making strong assumptions on how the world works.

@EikoFried A model with few covariates (eg L1) may be useful for etiology, but usually it will predict considerably worse than one with many small effects (L2), which, however, would be much harder to interpret 2/2

Are you attending @_useRconf next week? On Wednesday I will talk about #pencal, explaining how the package makes it easy to estimate and apply dynamic prediction models for datasets with numerous longitudinal covariates. Details here: sched.co/1c8uq #Rstats #useR2024

#RStats peeps: see you next week in Salzburg?

New (guest) blogpost on my blog: Hennig: "Statistical tests in 5 random research papers of 2024 & related thoughts on the don't say significance initiative" errorstatistics.com/2024/06/25/gue…"

@learnfromerror A great entry!

#Rstats package #pencal: version 2.2.2 is now on CRAN. It contains an updated vignette, which reflects all the changes introduced in version 2.2.1

They came to @EURO2024 from Ukraine - the defensive line of Europe.

Are you attending #UseR2024 next month? On Wednesday 10/7 I will give a talk on the #Rstats package #pencal, explaining how pencal makes it easy to estimate and apply dynamic prediction models for datasets with numerous longitudinal covariates. Details at sched.co/1c8uq

Source: projectpro.io/recipes/find-c… (point 3)

I'm sorry, what?!? #rstats

Even though this is over 2 years old, IMHO this still remains the best model extrapolation ever: "everyone in America will be gay by 2047" 😂