Paweł Lenartowicz
LightNews LightNews
banner
freestylerscientist.pl
Paweł Lenartowicz
@freestylerscientist.pl
450 followers 1K following 150 posts
Aspiring statistician and meta-researcher, sailor and activists.
#PhilosophyofScience
#OpenScience
#Uncertanity
Posts Replies Media Videos
freestylerscientist.pl
I tried prove that you are not correct, but Chat GPT went wrong immediately (lol).
August 11, 2025 at 3:02 PM
freestylerscientist.pl
So, here you are, I replicate exactly the same problem with empirical data from the FIRST open source article I found. (posting my browser history is sufficient?)

> power
[1] 0.6367
> p_beetween
[1] 0.1702

Source of data: www.nature.com/articles/s41... (excel in attached)
So, here you are, I replicate exactly the same problem with empirical data from the FIRST open source article I found. (posting my browser history is sufficient?) > power [1] 0.6367 > p_beetween [1] 0.1702 Source of data: https://www.nature.com/articles/s41467-024-55155-4/figures/1 (excel in attached)
February 9, 2025 at 11:03 PM
freestylerscientist.pl
I also mentioned it here: osf.io/preprints/me...
(I'll finish final version in a few days)
Another issue, which to my knowledge has not been widely discussed, is that p-curves — and consequently the z-values of some tests based on the t-distribution — do not always follow their theoretical distribution. This discrepancy arises when a t-test is applied to a heavy-tailed distribution, where individual data points can disproportionately impact the standard deviation estimate. In such cases, the distribution of t-statistics can shift toward values around 1 or -1. This characteristic, while making the t-test robust and conservative under assumption violations by preventing inflation of the error rate beyond the nominal $\alpha$-level (at the cost of statistical power), complicates interpretation near the significance threshold.
February 9, 2025 at 9:58 AM
freestylerscientist.pl
Here you are:
> power
[1] 0.5792
> p_beetween
[1] 0.22567

Code below, I've also posted it as a comment.
nSims <- 100000 #number of simulated experiments p <-numeric(nSims) #set up empty container for all simulated p-values n=20 simulation_chi_selective <- function(effect, n, ratio, df){ m = rbinom(1, n, ratio ) s2a = rnorm(n-m) s2b = rnorm(m) + rchisq(m, df = df) * effect s2 = append(s2a,s2b) return(s2) } for(i in 1:nSims){ #for each simulated experiment x<-rnorm(n = n, mean = 0, sd = 1) y<-simulation_chi_selective(effect = 2, n = n, ratio = 0.25, df = 8) z<-t.test(x,y) #perform the t-test p[i]<-z$p.value #get the p-value and store it } #now plot the histogram hist(p, main="Histogram of p-values", xlab=("Observed p-value"), breaks=40) #Observed Power power <- p[p<0.05] power <- length(power)/nSims p_beetween <- p[p<0.05 & p>0.025] p_beetween <- length(p_beetween)/nSims
February 8, 2025 at 4:23 PM
freestylerscientist.pl
Downplaying the impact—which is comparable in magnitude to that of moderate drinking during pregnancy—or calling for 'more research' before taking action is not "scepticism". It is a pseudoscientific tactic of 'manufacturing doubt,' akin to "smoking might be caused by cancer".
January 11, 2025 at 5:50 PM
freestylerscientist.pl
I do not understand. There IS OPEN SCIENCE MOVEMENT. It's about much more than "preregistration and open data" (even not everyone agree on principles).
January 10, 2025 at 2:50 PM
freestylerscientist.pl
There are 3 different common blue in many Slavic l. (for ex. Belarusian, Polish)

Гранатавы (Be), granatowy (PL) - dark blue
Сіні (Be), niebieski (PL) - blue
Блакітны (Be), błękitny (PL) - light blue

also, there are some regionalisms, as "modry" - Silesian, Czech, like "modra kapusta" (below)
"blue cabbage" or modra kapusta on picture.
December 14, 2024 at 7:50 PM
freestylerscientist.pl
I think that we could both agree that it's not a dilemma between "do not work at all" vs. "works almost as it should". But, again, in my opinion, the general problem is not "just-so-story". A decline in trust in science was predicted, see: "Scientific Knowledge and Its Social Problems"(1971) or
November 26, 2024 at 4:22 PM
freestylerscientist.pl
Most bias detection methods rely on assumptions that are not fully disclosed, tested or even addressed by critics. I have tried to describe why these methods should be used carefully. Thank you, @devezer.bsky.social and @jbakcoleman.bsky.social for persistently pointing this out.
Text show correction changing "questionable" to "different" research practices.
November 18, 2024 at 2:35 PM
freestylerscientist.pl
Mindless statistics as it's finest. 2 out of 3 significant results, in the absence of publication bias and an infinite number of potential errors, tell us that we should reject the H0 model and provide almost no information or evidence about the alternative hypotheses.
The null ritual: 1. Set up a statistical null hypothesis of “no mean difference” or “zero correlation.” Don’t specify the predictions of your research hypothesis or of any alternative substantive hypotheses. 2. Use 5% as a convention for rejecting the null. If significant, accept your research hypothesis. Report the result as p < 0.05, p < 0.01, or p < 0.001 (whichever comes next to the obtained p-value). 3. Always perform this procedure.
February 20, 2024 at 5:36 PM