Ruben C. Arslan

@ruben.the100.ci

4625 Followers 801 Following

Bayescurious evidence enthusiast at the100.ci Topics: evolution, ovulation, mutation, intelligence, personality, sexuality, R, open science & source tools. https://rubenarslan.github.io/

Suicide nets at Foxconn
The Chimpy Trinity
We also had concerns about the inferential leap from the video interview data to the paper's broader conclusions. Of 50 contacted low-quality respondents, 20 agreed to participate, and all 20 were human, a finding the authors use to support the claim that problematic respondents are generally not bots. However, automated bots cannot respond to a video interview invitation in the first place, so the sample is necessarily restricted to those capable of responding. The absence of bots among willing video interviewees does not tell us much about their prevalence among non-respondents. If 30 of the 50 contacted respondents were bots, such a result would also have been found.
screenshot of the R fiddle
Measuring self-citation using the patented 100% CI folding rule
Illustration of Berkson Error
Illustration of classical measurement error
Significance Testing after Model Selection Assume that the first step of investigating a (simple or complex) congruence hypothesis revealed that the respective hypothesis-conforming model did not fit the data significantly worse than the full third-order model. Then the coefficients of this model must be tested for significance in a second step, to determine whether the suggested effects are significant and in the expected direction. One open methodological issue refers to the question of how the model selection uncertainty from the first step can be considered when testing the coefficients in the second step, in order to avoid potential inflation of Type I error rates (Berk et al., 2010; Leeb & Pötscher, 2006). Different strategies have been suggested to account for such post-selection inference (e.g., splitting the data into a model selection sample and a coefficient testing sample, Berk et al., 2010; computing confidence intervals that consider all other models that might have been chosen by the model selection procedure, besides the one that was in fact chosen, Berk et al., 2013; or computing a confidence interval whose coverage probability is secured given the selected model, e.g., Lee et al., 2016; Rügamer & Greven, 2018). However, there seems to be no consensus yet on which method is the "most valid" one. Thus, we cannot yet make a well-informed recommendation on how to account for post-selection inference in cubic RSA. In the illustrative R Code, we decided to reduce the alpha level to .01, but this CUBIC RSA 52 approach should be seen as preliminary and should be adapted on the basis of future methodological work on the question of post-selection inference.
me being mildly cross at best about a low ceiling
The evolutionary biologist Richard Dawkins made the point that ge­ ne­ tic ­ causes should be defined as difference makers even for relatively ­ simple pheno- types that are intuitively “ge­ ne­ tic,” such as eye color. He wrote, “The ‘effect’ of any would-be cause can be given meaning only in terms of a comparison, even if only an implied comparison, with at least one alternative cause. It is strictly incomplete to speak of blue eyes as ‘the effect’ of a given gene G1. If we say such a ­ thing, we ­ really imply the potential existence of at least one alternative allele, call it G2, and at least one alternative phenotype, P2, in this case, say, brown eyes.” He continues with an example of two genes both related to skin pigmentation: “To be sure, A, the gene whose protein product is the black pigment, is necessary in order for an individual to be black. . . . But I ­ shall not call A a gene for blackness ­ unless some of the variation in the population is due to lack of A. . . . The point that is relevant ­ here is that both A and B are potentially entitled to be called genes for blackness, depending on the alternatives that exist in the population (emphasis
Since its inception, the eld of evolutionary psychology has been accused of such lack of rigor by critics who often assert that evolutionary psycho- logical hypotheses cannot, in principle, be tested empirically (e.g., Gould & Lewontin, 1979; Rutherford, 2020). In this
Arguments that the presence of brain disorders at high frequency in Ashkenazi Jews might explain the enrichment of genes that boost brains is woolly conjecture, and can be abandoned with current data to hand. “Winnowing through persecution” is also merely idle speculation, and has no place in a decent scientific paper. These are fractionally more sophisticated versions of the evolutionary crime we call adaptationism*—the assumption that natural selection is responsible for specific human behaviors, rather than happenstance or processes that are neither positive nor negative, but have simply drifted into existence. In the genomic age we are capable of actually seeing the parts of the genome where selection has taken place, and there are population-specific mutations that indicate positive selection of particular genes as adaptations to the local environment. Pigmentation, specific diets, resistance to diseases (such as malaria), and other traits are demonstrably local adaptations that are part of humankind’s success in colonizing the world. Adaptationism is an error because in many cases it results in untestable hypotheses, but ones that are appealing because they sound superficially convincing—Blacks are good sprinters because of selection during slavery; Jews are intellectually gifted because their history of persecution enriched
violin plot of results by testosterone measure
“First, parenting matters. Much more consistent than any of the associations in these studies is the association between parenting and child outcomes. Having books in your house and reading them to your kid is going to matter much more than what books they have at day care. This seems to be true even though your child probably spends as many waking hours with their care providers as with you. I don’t think we know precisely why this is the case, although it may be that you as the parent are the most consistent influence your child has.”
Markus Steiner exploding balloons for science.
the golden area is the prediction interval.
Fig. 3. Scatter plots of the synthetic and empirical estimates, validation study (Stage 2). Showing N = 30,135 item-pair correlations, N = 257 scale reliabilities, and N = 1,568 scale-pair correlations for (top) the pretrained SBERT model and (bottom) the fine-tuned SurveyBot3000 model. SBERT = all-mpnet-base-v2 model.
Plot from a master thesis I supervised, building on Willoughby et al. Lefthandedness, breast cancer, and sexual orientation stand out as traits where lay estimates for being mainly genetic are much larger than twin study estimates.
The Galton board in which many small influences sum to a normal distribution.
Stabilities of intraindividual Emotional Variability Note. Stabilities of emotional variability, estimated on latent and manifest levels, separately for negative (A) and positive emotion (B) and each data set (Uni2work and COVID data sets). The latent correlations shown are from the two-wave BCLSMs; the manifest correlations are from the separate, one-wave BCLSMs. Each estimation is accompanied by its corresponding 95% credible interval in brackets.
‘A fragmented field: Construct and measure proliferation in psychology.’ (2025)
Fig. 3. Effects of the intervention. In (a) are shown effects 7 months after the intervention; in (b) are shown interaction effects 7 months after the intervention. Models show the effects of the intervention on expressive-language skills in the grade 1 posttest, with an interaction between expressive-language skills in the pretest and after the intervention. Standardized coefficients (with 95% confidence intervals) are shown, except for the intervention dummy variable and the interaction where y-standardized values are indicated. Rectangles and circles contain observed variables and latent variables, respectively. Solid arrows point to significant regression or factor loadings (arrows from latent variables to their observed indicators); the arrow with dotted lines shows nonsignificant regression. The interaction (Fig. 3b) is illustrated by the arrow from the black circle to expressive language in the posttest. **p < .01.
bf_snp <- bf(WGS_h2 | se(WGS_h2_SE, sigma = TRUE) ~ 1) bf_twin <- bf(Twin_Heritability ~ 1) fit_ratio <- brm( bf_snp + bf_twin, data = dat %>% filter(Trait != "" & !is.na(Trait)), chains = 4, iter = 4000, warmup = 1000 ) # Extract draws and compute ratio draws <- as_draws_df(fit_ratio) # Check variable names # get_variables(fit_ratio) ratio_draws <- draws$b_TwinHeritability_Intercept / draws$b_WGSh2_Intercept # Summary data.frame( Estimate = mean(ratio_draws), Est.Error = sd(ratio_draws), Q2.5 = quantile(ratio_draws, 0.025), Q97.5 = quantile(ratio_draws, 0.975) )
> brms_eiv Family: gaussian Links: mu = identity Formula: Twin_Heritability ~ me(WGS_h2, WGS_h2_SE) Data: dat %>% filter(Trait != "" & !is.na(Trait)) (Number of observations: 41) Draws: 4 chains, each with iter = 4000; warmup = 1000; thin = 1; total post-warmup draws = 12000 Regression Coefficients: Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS Intercept 0.44 0.05 0.35 0.53 1.00 4362 5107 meWGS_h2WGS_h2_SE 0.44 0.16 0.13 0.75 1.00 4307 5781 Further Distributional Parameters: Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS sigma 0.12 0.01 0.10 0.15 1.00 5872 6274
Funnel plots from Polderman et al. 2015
Amazingly, with three different method and datasets, the estimated “narrow-sense” heritability comes in at ~30% (shown above in blue). What about rare variants? In the two studies where common variant / GWAS heritability was estimated (Young et al. and Wainschtein et al.), it accounted for ~85% of the total narrow-sense heritability estimate, indicating that rare variants make a relatively small contribution on average and GWAS provides an only slightly lower bound on total narrow-sense heritability. As tentative support, Wainschtein et al. conducted an exploratory analysis of the heritability from 760 million “ultra-rare” variants and found that they added essentially nothing (~0.12%) to the total heritability estimate on average.
After adjusting for geographic clustering the heritability estimates dropped further (38-40%). After adjusting for more geographic clustering, the heritability estimates dropped even further (32-34%). As a comparison, we can look at the same covariate adjustment for standing height and see that there is essentially no impact whatsoever. In fact, education and IQ exhibited by far the strongest evidence of stratification compared to the other traits analyzed (which were largely anthropometric or blood/lipid-oriented). GREML-WGS heritability estimates for three traits with a variety of covariates. Data from Wainschtein et al. (2025) The authors stopped at 100 geographic clusters, but does this adjustment fully correct for environmental biases (or maybe over-correct?) — no one really knows! We do know that a substantial fraction of the apparent educational attainment “heritability” is actually indirect associations from stratification or parental effects, and that disentangling these associations requires genotyped family data. So in a fundamental sense, GREML-WGS is the wrong tool for this question (direct narrow-sense heritability) for these traits (stratified behavior)4.
Box 1 — A Case Study in LLM Pollution Detection and Incidence In a recent study we conducted with participants recruited via Prolific, we found that a lack of bot protection measures led to an alarming rate of seemingly bot-generated answers. This example is used here illustratively to demonstrate the types of challenges that can arise in practice rather than to report new empirical findings. In our initial pilot study, we only tracked copying and pasting on a page containing an openended question. Notably, 45% of participants engaged in one or both of these actions, suggesting potential cases of Partial LLM Mediation and Full LLM Delegation. Even to an untrained eye, some of these responses were obviously generated by LLMs, representing, for instance, cases of overly-abstract summarisations of all previous instructions. Given these results, we implemented the following measures for the main study: • Implemented two reCAPTCHA v2 checkpoints at the beginning and middle of our survey, asking participants to confirm that they are not bots by checking a box and potentially completing a short validation test (0.2% failed to advance); • Activated Qualtrics’ version of reCAPTCHA v3, which uses information about users’ behaviour and past activity to assign each a score between 0 and 1, with lower scores indicating likely bot activity (2.7% participants scored lower than 0.7, a threshold we determined on the basis of our piloting, with the lowest score being 0.2); • Added a honeypot question that included white text invisible to humans, but readable to bots scraping all content (1.6% participants included our study-irrelevant keyword “hazelnut”; out of these 16 participants, only 2 had low reCAPTCHA v3 scores); • Prevented copy-pasting, while still tracking which participants attempted these actions (4.7% participants attempted to copy and/or paste text). While these incidence rates were less concerning than those observed during piloting, we still identified identical nonse…
Simulated data showing a spurious interaction arise due to a ceiling effect.
me with some garden hoses connected in a X -> Z <- Y fashion. If I shut the valve at Z, water from X spills out at Y
> library(tidyverse) > > squish <- function(x) { + x <- round(x) + x[x>7] <- 7 + x[x<1] <- 1 + x + } > nonserious <- 0.1 > N <- 500 > dt <- tibble( + serious = sample(c(F, T), N, prob = c(nonserious, 1-nonserious), replace = T), + poop = if_else(serious, 1, sample(1:7, N, T)), + chocolate = if_else(serious, squish(rnorm(N, 6.5, sd = 0.5)), sample(1:7, N, T)), + ) > > table(dt$serious) FALSE TRUE 44 456 > hist(dt$poop) > hist(dt$chocolate) > > effsize::cohen.d(dt$poop, dt$chocolate, paired = T) Cohen's d d estimate: -4.866682 (large) 95 percent confidence interval: lower upper -5.409463 -4.323901
A fragmented field. If we conceive of the constructs and concepts studied in behavioral science as a map, we would find that it is highly fragmented and directions are hard to come by. Scientists can hardly stand on each others' shoulders if they cannot manage to meet on common ground.
but it might work for us
Bubble plot
We provide the first evidence from a randomized double-blind placebo control study of oral contraceptives on women's preferences for masculine and symmetric faces in men. We find no evidence of an effect. Similar, our correlational results testing for an association between menstrual cycle phases and facial preferences fail to reject the null hypothesis of no association, as do our estimated correlations between hormone levels and facial preferences. Although the early literature largely found support for variation in facial preferences over the menstrual cycle and with oral contraceptive use (Feinberg et al., 2008; Johnston et al., 2001; Little et al., 2002; Little et al., 2013; Penton-Voak et al., 1999; Roney & Simmons, 2008), our null results align with more recent large-sample correlational studies (Jones et al., 2018; Marcinkowska et al., 2019). One possible motivation of the hypothesis that oral contraceptives lower the preferences for facial masculinity and symmetry is in terms of the dual mating theory. In a recent review of the literature Jones et al. (2019) point out several methodological problems in the early work testing this hypothesis, such as low statistical power, and conclude that recent and methodologically stronger work fails to support the hypothesis. Our findings lend further support for their conclusions. There are also several reasons for why false positive results could have been published in the earlier literature. Low statistical power increases the probability for false positive results as well as inflated effect sizes. Many of the older studies also rely on self-reported menstrual cycle data, which is subject to mismeasurement. In addition, without pre-analysis plans there is room for the so-called “researcher degrees of freedom” (Simmons et al., 2011) to influence the analysis such that statistically significant results become more likely to be false positive ones. This can occur through excluding observations, using covariates, testing…
Treemap showing fragmentation across industrial/organizational psychology. Fragmentation (Hill-Shannon Diversity) 𝐷=671.
Normalised Shannon entropy. Dotted pre-correction, normalised using number of tests, solid after, normalised using number of distinct tests.
measurement fragmentation standardized using fragmentation, with different hill numbers
Here, we report results across years in the Hill-Shannon diversity metric (exponentiated Shannon entropy) using asymptotic coverage standardization. The Hill-Shannon diversity is more interpretable than Shannon entropy because changes in it are linear in the unit of number of measures/constructs. If the Hill-Shannon diversity for a subfield is 100, it can be understood as saying that the subfield has the same fragmentation as an imaginary subfield with 100 tests that are all used equally often. Hill-Shannon diversities can be compared relatively, so a subfield with a value of 200 can be said to be twice as fragmented as a subfield with a value of 100.
Number of distinct measures.
APA upped its standards in 2016, thereafter more measures had complete information about validity, factor structure etc.
Treemap showing measurement fragmentation across subfields in psychology. Hill-Shannon Diversity 𝐷=1626.05