Konsta Happonen @konsta.happonen.eu · 25/08/2026Thanks for the great suggestions @tiddlydump.bsky.social ! The food at Etno Dvaras was delicious, and the gardens were beautiful. It was a nice surprise to find a specialty coffee joint right at the edge of the gardens with a view to the systematic section. 220
Konsta Happonen @konsta.happonen.eu · 24/08/2026Off to #ESAWarsaw2026. Boat ride from Helsinki to Tallinn done, now on the train to Vilnius. Tomorrow the final train to Warsaw! Good thing I brought something to pass the time with. #statssky 2121
Konsta Happonen @konsta.happonen.eu · 22/08/2026I'm unironically in favor of changing the required format of publications from Word documents and pdfs to plain text files. 010
Konsta Happonen @konsta.happonen.eu · 30/07/2026I have convinced my partner that we should start playing Blood Bowl, but their condition is that they get to play as the Gnomes. It might be that we will both have to play Gnomes. Look at these minis!!! (Especially the fox with the ball and the goose). 020
Konsta Happonen @konsta.happonen.eu · 25/07/2026Probably the last one I'm finishing before summer vacation is over. I'm quite happy with how the color gradations on the skin and the loincloth turned out with this one! 030
Konsta Happonen @konsta.happonen.eu · 27/06/2026One of my favourite new things here is the Kruunuvuorensilta bridge. The longest in Finland, and it has exactly zero lanes for cars: only pedestrians, cyclists and trams allowed. 230
Konsta Happonen @konsta.happonen.eu · 21/06/2026Deep one coming along. Painting minis is such a nice way to turn your brain off. It requires just enough focus that your mind can't really wander, but not too much to feel exhausting. 130
Konsta Happonen @konsta.happonen.eu · 10/06/2026For example, I learned that there's a class of distributions where the existence of moments depends on the value of a parameter. 110
Konsta Happonen @konsta.happonen.eu · 26/05/2026Much of the game is orchestrated by the managers, who communicate to the batters and the runners using a multi-coloured fan. 100
Konsta Happonen @konsta.happonen.eu · 26/05/2026What makes offence hard is the relatively compact form of the field combined with the backline: there usually isn't a lot of space far away from defence players. If a defence player catches the ball before it hits the ground, all advancing runners have to return home. 100
Konsta Happonen @konsta.happonen.eu · 26/05/2026These changes increase the importance of strategy and de-emphasize the pitcher-batter duel. Hitting the ball is relatively easy — what counts is whether the batter can place the ball where they want. 100
Konsta Happonen @konsta.happonen.eu · 26/05/2026Four of the biggest changes compared to baseball are 1) The pitcher pitches directly upward from the home base 2) The field is more compact and the bases in a zig-zag formation. 3) Hitting the ball over the backline results in a foul ball. 4) A fair hit does not force runners to advance. 110
Konsta Happonen @konsta.happonen.eu · 26/05/2026The story goes that a Finnish guy Tahko Pihkala travelled to the United States in the early 1900s, saw a game of baseball, and brought it with him back to Finland. However, he thought the game could be improved in a couple of ways. 100
Konsta Happonen @konsta.happonen.eu · 26/05/2026Need a break from work, so y'all are getting a thread about the universally best sport there is: PESÄPALLO, or Finnish baseball. 111
Konsta Happonen @konsta.happonen.eu · 18/05/2026Growing old so it's time to return to all the hobbies I had when I was a teenager, such as painting small plastic people. I'm very happy with the model on the left, considering it's the first mini I've painted in 20 years. 120
Konsta Happonen @konsta.happonen.eu · 29/04/2026Oh, OK. Both these examples seem to run OK, but again, I have no idea whether what marginaleffects is doing under the hood is the correct thing. I think I'll try to run some more thoughtful simulations later. 000
Konsta Happonen @konsta.happonen.eu · 28/04/2026This is cool. @tslumley.bsky.social implemented a predict method for svyVGAM::svy_vglm models, which means that @vincentab.bsky.social's marginaleffects can now in theory be extended to accomodate these models. #rstats 2181
Konsta Happonen @konsta.happonen.eu · 28/04/2026Never put one of these on my side monitor before, but damn it's relaxing to have an owl nest on your desk. wwf.fi/en/wildlive/... 110
Konsta Happonen @konsta.happonen.eu · 18/04/2026Crazy coincidence that just the night I'm in Stockholm it's Kulturnatten, so I got to listen to Renaissance and Baroque vocal music in gorgeous Katarinakyrkan. 030
Konsta Happonen @konsta.happonen.eu · 04/04/2026Sounds like a perfect combo with Swedish banana curry pizza. 000
Konsta Happonen @konsta.happonen.eu · 03/04/2026Most of the recipes come from Hannah Che's The Vegan Chinese Kitchen, which I can highly recommend. 030
Konsta Happonen @konsta.happonen.eu · 03/04/2026Spent hours today preparing these, but it was worth it. I'm especially happy to have discovered smashed cucumber salads. This one was made with finely chopped umeboshi and sesame oil. The fresh crispiness of the cucumbers goes so well with the tart saltiness of the umes. 170
Konsta Happonen @konsta.happonen.eu · 02/04/2026I think something like this example from the legendry package is probably less distorting. teunbrand.github.io/legendry/art... 030
Konsta Happonen @konsta.happonen.eu · 31/03/2026This time on #3DChartsInTheWild: a 3D bar chart from a published journal article. 130
Konsta Happonen @konsta.happonen.eu · 25/03/2026Data quality looking OK. Pink: proportion of non-missing non-can't say responses per respondent. Brown: Can't say, black: Missing answer. About 10 % failed an instructed response item. Is that good or bad? At least it's better than 60 % (doi.org/10.1177/0049...). 110
Konsta Happonen @konsta.happonen.eu · 23/03/2026Googling about different interpretations of the character of Pontius Pilate among Christians. So happy that we now have this incredible technology to help us in our quest for knowledge. 180
Konsta Happonen @konsta.happonen.eu · 23/03/2026I guess it depends on what you prefer! Personally I like February, because it's one of the months when we're most likely to have proper winter weather with dependable subzero temperatures. That makes it possible to do stuff like walk on sea ice, go ice skating etc. 140
Konsta Happonen @konsta.happonen.eu · 17/03/2026Uh oh. I guess it's time to do that thing that's indistinguishable from data falsification. 121
Konsta Happonen @konsta.happonen.eu · 12/03/2026The second best thing is that the Finnish Easter delicacy mämmi is in season again. 020
Konsta Happonen @konsta.happonen.eu · 09/03/2026As a non-native speaker, I disagree. The journal boasts that it gives a highly respected home for research, yet cannot be bothered to check even the first sentence of the article for coherence. It costs 2500€ + VAT to publish in Scientific Reports. I absolutely think that warrants ridicule. 130
Konsta Happonen @konsta.happonen.eu · 09/03/2026My favorite part from this piece is "This part show how change in number make line in graph up or down". 100
Konsta Happonen @konsta.happonen.eu · 09/03/2026Today I discovered that listening to plainchant while working seems to help with stress. There's something very comforting about it, especially the pieces with a low drone in the bass. 010
Konsta Happonen @konsta.happonen.eu · 25/02/2026I want to be able to do stuff like this, but I'm told it's not a good idea in #rstats >:| 020
Konsta Happonen @konsta.happonen.eu · 25/02/2026Thing that goes up; Thing that goes down; Thing that stays about the same; Secret fourth thing; 100
Konsta Happonen @konsta.happonen.eu · 20/02/2026I mostly love tibbles, especially because of pretty printing, but this should be illegal. 490
![library(tidyverse)
library(survey)
library(srvyr)
library(svyVGAM)
library(marginaleffects)
dd <- tibble(
y = sample(LETTERS[1:5], 100, replace = TRUE),
x = rnorm(100),
weights = 1 + abs(rnorm(100, 0, 0.1))
)
dd <- dd |> as_survey_design(weights = weights)
m1 <- svy_vglm(y ~ x, multinomial(refLevel = "A", ynames = TRUE), dd)
source("R/marginaleffects_extension.R")
m1 |>
avg_comparisons(
newdata = datagrid(x = 0),
variables = "x",
type = "response"
)](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:5muznqvsk4hath65gqiapmej/bafkreibgc2k4j3wbybt3wpl5yvj2kmaiap7szdfcxh6iwv2l72ff6xrxxq)
![tib[,1] returns a tibble instead of a vector.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:5muznqvsk4hath65gqiapmej/bafkreiefquilc3rbcq3cp4dz5t6sqnh3gdlg4zw2o7n7hhygsq5u7yrge4)
![If [some variables] are an IID sample from the standard Cauchy distribution, then their sample mean is also standard Cauchy distributed. In particular, the average does not converge to the mean, and so the standard Cauchy distribution does not follow the law of large numbers.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:5muznqvsk4hath65gqiapmej/bafkreiay6bftuhdohwpl2atdoxo2jth6eoxxyt75j42vah3gtthagaxkl4)
