Waschi Waschoi
@waschoi.bsky.social
background: previously long, linear routines that were easy to debug, similar to statistical scripts. Refactoring repeated code into functions (no more than twice” rule) improved structure but made debugging and development feel harder. deeper understanding of function design and workflow.
January 5, 2026 at 8:32 PM
background: previously long, linear routines that were easy to debug, similar to statistical scripts. Refactoring repeated code into functions (no more than twice” rule) improved structure but made debugging and development feel harder. deeper understanding of function design and workflow.
Package development: focus on function design and structure—deciding when to create new functions or reuse internal routines, when to apply rlang and tidy evaluation (dots), and when to define explicit arguments. Also includes debugging strategies in RStudio and key literature references.
January 5, 2026 at 7:59 PM
Package development: focus on function design and structure—deciding when to create new functions or reuse internal routines, when to apply rlang and tidy evaluation (dots), and when to define explicit arguments. Also includes debugging strategies in RStudio and key literature references.
This was so nice of you! Thank you 🙏
December 26, 2025 at 2:08 PM
This was so nice of you! Thank you 🙏
This is so usefull. 🙏🙏🙏
December 19, 2025 at 10:35 PM
This is so usefull. 🙏🙏🙏
Can do it, but there are a lot of great and better ressources. But I made a nice solution with AIs help which I can share later today. Solved the problem of qwen 2.5vl getting stuck in an ocr loop. Took the Model over 10 minutes to extract some text which is an absolute no go
November 27, 2025 at 3:36 AM
Can do it, but there are a lot of great and better ressources. But I made a nice solution with AIs help which I can share later today. Solved the problem of qwen 2.5vl getting stuck in an ocr loop. Took the Model over 10 minutes to extract some text which is an absolute no go
www.gilesd-j.com/2025/01/20/f... seems like a good start. But did not find real advantages compared to Öllama. Maybe I missed something
From code to conversation: Localized AI fun with LM Studio and the ellmer package – Giles
www.gilesd-j.com
November 27, 2025 at 2:56 AM
www.gilesd-j.com/2025/01/20/f... seems like a good start. But did not find real advantages compared to Öllama. Maybe I missed something
Do you also have any experience with Ollama? Does LLMStudio offer any advantages? I would then be very interested in example code for interacting with Ellmer. Perhaps you know of a good resource?
November 27, 2025 at 2:43 AM
Do you also have any experience with Ollama? Does LLMStudio offer any advantages? I would then be very interested in example code for interacting with Ellmer. Perhaps you know of a good resource?
Are you using ollmam for the local llms? Which model do you use for this task? Had good outcomes with qwen3 at similar tasks.
November 27, 2025 at 2:32 AM
Are you using ollmam for the local llms? Which model do you use for this task? Had good outcomes with qwen3 at similar tasks.
str_flatten can handle na.
November 26, 2025 at 12:11 PM
str_flatten can handle na.
No it allows real weights. I normaly use expss::w_median which uses matrixstats. The results were identical with DescTools. Now I use collapse and could reproduce my previous analysis but your blog post made me aware of potential flaws.
November 5, 2025 at 7:23 PM
No it allows real weights. I normaly use expss::w_median which uses matrixstats. The results were identical with DescTools. Now I use collapse and could reproduce my previous analysis but your blog post made me aware of potential flaws.
That's really exciting. Since many of my colleagues still use SPSS, I always have to make sure that R and SPSS produce identical results. I'll run a simulation myself.
November 5, 2025 at 5:09 PM
That's really exciting. Since many of my colleagues still use SPSS, I always have to make sure that R and SPSS produce identical results. I'll run a simulation myself.
Use annote. It is an extra text layer
ggplot2.tidyverse.org/reference/an...
ggplot2.tidyverse.org/reference/an...
Create an annotation layer — annotate
This function adds geoms to a plot, but unlike a typical geom function,
the properties of the geoms are not mapped from variables of a data frame,
but are instead passed in as vectors. This is useful ...
ggplot2.tidyverse.org
November 5, 2025 at 11:23 AM
Use annote. It is an extra text layer
ggplot2.tidyverse.org/reference/an...
ggplot2.tidyverse.org/reference/an...
Thanks so much for this great solution and all your work!
November 4, 2025 at 12:49 PM
Thanks so much for this great solution and all your work!
Had the same experience. At the beginning it felt great and I made fast progress. But now I have to refactor everything and lost control 🙈 but it was a nice learning
November 1, 2025 at 1:46 PM
Had the same experience. At the beginning it felt great and I made fast progress. But now I have to refactor everything and lost control 🙈 but it was a nice learning
For me it is always the proxy at my company. Sys.setenv(http_proxy='xxx.com:80',https_proxy='xxx.com:80') might help.
October 21, 2025 at 3:57 PM
For me it is always the proxy at my company. Sys.setenv(http_proxy='xxx.com:80',https_proxy='xxx.com:80') might help.
It can handle missing values and has a lot of options but is not overloaded
October 7, 2025 at 5:46 PM
It can handle missing values and has a lot of options but is not overloaded
autumn is really good!
github.com/aaronrudkin/...
github.com/aaronrudkin/...
GitHub - aaronrudkin/autumn: autumn: Fast, Modern, and Tidy-Friendly Iterative Raking in R.
autumn: Fast, Modern, and Tidy-Friendly Iterative Raking in R. - aaronrudkin/autumn
github.com
October 7, 2025 at 5:44 PM
autumn is really good!
github.com/aaronrudkin/...
github.com/aaronrudkin/...
Auch wenn ich den Ärger verstehen kann: ohne Mietspiegel und Kappungsgrenze wären Mieter/-innen mit ganz anderen Forderungen konfrontiert.
September 19, 2025 at 8:23 AM
Auch wenn ich den Ärger verstehen kann: ohne Mietspiegel und Kappungsgrenze wären Mieter/-innen mit ganz anderen Forderungen konfrontiert.
Das ist natürlich Unsinn. Der Mietspiegel bildet die durchschnittliche Miete realistisch ab. Die Datenbasis ist mit 5999 Fällen deutlich oberhalb der gesetzlichen Vorgaben. Interessant ist immer, dass es den Mietern zu teuer und den Vermietern im Ergebnis zu günstig ist.
September 19, 2025 at 8:23 AM
Das ist natürlich Unsinn. Der Mietspiegel bildet die durchschnittliche Miete realistisch ab. Die Datenbasis ist mit 5999 Fällen deutlich oberhalb der gesetzlichen Vorgaben. Interessant ist immer, dass es den Mietern zu teuer und den Vermietern im Ergebnis zu günstig ist.
OK. Dann würde ich das mal als realistischeres Bild im Vergleich zum 2022er interpretieren. Aber an der Situation der zu hohen Miete in Vergleich ändert das ja nichts. Erhöhen kann dein Vermieter dann aber entsprechend nicht. Schönen Abend noch und Danke für die Rückmeldung
September 18, 2025 at 7:38 PM
OK. Dann würde ich das mal als realistischeres Bild im Vergleich zum 2022er interpretieren. Aber an der Situation der zu hohen Miete in Vergleich ändert das ja nichts. Erhöhen kann dein Vermieter dann aber entsprechend nicht. Schönen Abend noch und Danke für die Rückmeldung