Lightnews — Scholar-powered news

Juli Nagel

@juli-nagel.bsky.social

180 followers 130 following 83 posts

PhD student (psychology, neurobiology) researching reward, memory and sleep. Way too passionate about R. Otherwise interested in gaming, western riding and climbing. Mum.

Posts Media Videos Starter Packs

Juli Nagel @juli-nagel.bsky.social · 7d

Caveat: tabulate() only works for single vectors, so you can't replace sth like table(1:4, c("a", "a", "b", "b")).
Bonus points: You easily get 0 counts if you expect numbers from 1:nbins, i.e.:

> tabulate(c(3, 1, 3, 3, 3), nbins = 4)
[1] 1 0 4 0

2/2

Juli Nagel @juli-nagel.bsky.social · 7d

IF YOU CAN, REPLACE TABLE() WITH TABULATE() IN YOUR #rstats CODE!

data <- sample(1:3, 100, replace = TRUE)
microbenchmark::microbenchmark(table(data), tabulate(data), times = 10000)

I'm almost angry at how much faster this is, and why I was today years old when I learned this 🤬
1/2

Juli Nagel @juli-nagel.bsky.social · 8d

This is extremely exciting!
(But also with the release party being on October 3rd, a missed opportunity to use a "The Life of a Showgirl" reference.)

Juli Nagel @juli-nagel.bsky.social · 9d

Is this ... a very early preregistration?

Juli Nagel @juli-nagel.bsky.social · 9d

I also learned a fair deal of Spanish in school (read Don Quijote and all), but couldn't even do small talk nowadays. That happens, and it's normal (albeit sad). I fell into the "What?! But EVERYBODY should know THIS!" trap in the past, but learned that it's a very unhelpful trap to fall into.
2/2

Juli Nagel @juli-nagel.bsky.social · 9d

Sure, I've learned this in school. I'm a nerdy nerd with degrees in psychology, working as a scientist, well-trained in statistics (not so much physics). When I see these videos, all I remember is that it's not surprising that this happens, but I wouldn't be able to explain it properly.
1/2

Juli Nagel @juli-nagel.bsky.social · 9d

To be fair, I've had occasional situations like these with students BEFORE the rise of LLMs ...

Juli Nagel @juli-nagel.bsky.social · 12d

Line 1: warning("Hey, thought about restarting R before running this script? Btw, you might wanna set up a clean project for this, cause, you know, if you still use setwd(), Jenny Bryan will personally set your computer on fire. And do you know about dependency management? Sorry if you already do!")

Juli Nagel @juli-nagel.bsky.social · 19d

Don't know why the zebra paper won, though? It's been tested in horses before (Caro et al.); the horse community has been using blankets with zebra stripes for years. Sure, we desparetly need more replications, but why the Ig Nobel prize for a conceptual replication?
www.science.org/content/arti...

Zebra stripes confuse biting flies, causing them to abort their landings

Coat pattern helps zebras and horses evade dangerous insects

www.science.org

Juli Nagel @juli-nagel.bsky.social · Aug 29

Had the opposite experience lately. Daycare warned me: "In case tells any stories about today ... don't be confused ... he was very excited about the document shredder."
Unfortunately, he DIDN'T mention it, so I will never know why they were so worried what weird things he might say.

Juli Nagel @juli-nagel.bsky.social · Aug 25

Wooooooow, where is this?! 😱

Juli Nagel @juli-nagel.bsky.social · Aug 20

"So I leap from the gallows and I levitate down your street
Crash the party like a record scratch as I scream
'Who's afraid of little old me?'
You should be"

Juli Nagel @juli-nagel.bsky.social · Aug 20

(Disclaimer: Of course, results will vary if the data are not balanced, or are funky in any other way, but in this simple case, mixed model/t-test estimates and p-values will be identical.)

Juli Nagel @juli-nagel.bsky.social · Aug 20

t_test <- t.test(t1, t2, paired = TRUE)
mm <- lmer(value ~ time + (1 | id), data = data)
mm_summary <- summary(mm)

# estimates
abs(mm_summary$coefficients["timet2", "Estimate"])
abs(t_test$estimate)

# pval
mm_summary$coefficients["timet2", "Pr(>|t|)"]
t_test$p.value

2/2

Juli Nagel @juli-nagel.bsky.social · Aug 20

And here's a paired t-test versus a mixed model with a random intercept in R:

library(lmerTest)

n <- 1000
t1 <- rnorm(n)
t2 <- t1 + rnorm(n)

data <-
data.frame(
id = paste0("ID-", 1:n),
t1 = t1,
t2 = t2
)

data <-
data %>%
pivot_longer(c(t1, t2), names_to = "time")

1/2

Juli Nagel @juli-nagel.bsky.social · Aug 20

I once had a student telling me that she hates t-tests, but prefers linear mixed models. Boy, did I have news for her ...

Juli Nagel @juli-nagel.bsky.social · Aug 15

"Nagel" ist German for "nail", which is why this is the first GIF to pop up, once again reminding me how lucky I was to be able to assimilate my husband's last name.

Juli Nagel @juli-nagel.bsky.social · Aug 14

I am very proud of my Bluesky feed, which combines magic Quarto plugins with "The Life of a Showgirl" references.

Emil Hvitfeldt @emilhvitfeldt.bsky.social · Aug 13

Excited to share my newest quarto revealjs plugin: imagemover

Easily reposition and resize images directly in your quarto revealjs slides for a much smoother slidecrafting experience

github.com/EmilHvitfeld...
#quarto

Juli Nagel @juli-nagel.bsky.social · Jul 23

E.g.
www.journals.uchicago.edu/doi/abs/10.1...

(But personally, I think it's not either/or, and even IF base rates were the main explanation, we should STILL fight against questionable research practices.)

Understanding the Replication Crisis as a Base Rate Fallacy | The British Journal for the Philosophy of Science: Vol 72, No 4

The replication (replicability, reproducibility) crisis in social psychology and clinical medicine arises from the fact that many apparently well-confirmed experimental results are subsequently overtu...

www.journals.uchicago.edu

Juli Nagel @juli-nagel.bsky.social · Jul 23

Joke aside, we recently discussed in our journal club about how there's no nominal rate of successful replication, i.e., the replication rates we're seeing might be just what to expect. I don't buy that, BUT there's an interesting discussion here about base rates of theories being correct etc. etc.

Juli Nagel @juli-nagel.bsky.social · Jul 17

I think it says a lot about how dominant RStudio is as an IDE that people frequently mistake it for the actual programming language. Sure, many don't even think about the difference between a language and an IDE, but you'd never see someone write "we used Spyder for data analysis" (I hope).

Juli Nagel @juli-nagel.bsky.social · Jul 15

Also, I wonder what these data would look like if they were normaized to account for increasing overall publication rates. Here's a banana for scale! 🍌
arxiv.org/abs/2102.06418
@dalmaijer.bsky.social

Banana for scale: Gauging trends in academic interest by normalising publication rates to common and innocuous keywords

Many academics use yearly publication numbers to quantify academic interest for their research topic. While such visualisations are ubiquitous in grant applications, manuscript introductions, and revi...

Juli Nagel @juli-nagel.bsky.social · Jul 10

🖖

Juli Nagel @juli-nagel.bsky.social · Jul 8

I can see past the questionable artistic choices if they can also give me the data normalized by the overall consumption of each food item.

Reposted by Juli Nagel

Simon Kern @skjerns.de · Jun 16

preprint alert 🚨
1/ Can we accurately detect sequential replay in humans using Temporally Delayed Linear Modelling (#TDLM)? In our recent study, we could not find any replay and decided to dig deeper by running a hybrid simulation with surprising results. Link to preprint & details below 👇