Taylor et al. respond to recent #commentary with a re-run of their #analysis, with no #significant change to their #oddsratio

www.bjanaesthesia.org/article/S0007-0912(25)00...

U01.06.005 Quantifying risk Analyze the study design carefully. Start by identifying how participants were selected — note that cases were chosen based on disease status (neurological cancer) and controls were chosen based on being disease-free. Next, compare exposure rates (solvent exposure) between the two groups. This approach helps determine whether the exposure is associated with disease occurrence. Finally, recognize the type of epidemiological study and calculate the appropriate measure of association.

Quantify disease risk using relative risk, odds ratio, and attributable risk in epidemiologic analysis. #QuantifyingRisk #RelativeRisk #OddsRatio #AttributableRisk #Epidemiology #Biostatistics #EvidenceBasedMedicine #PublicHealth #MedicalEducation #USMLE

Demetri Pananos Ph.D - You Just Said Something Wrong About Logistic Regression

#statstab #321 You Just Said Something Wrong About Logistic Regression by @PhDemetri

Thoughts: Odd, probabilities, and risk ratios. Coefficients in logistic regression are only one of these.

#logisticregression #oddsratio #riskratio #probability #odds

dpananos.github.io/posts/2024-0...

#statstab #206 Probability, log-odds, and odds

Thoughts: A great short document covering the differences between the three, and some R code snippets on how to compute and convert them.

#logodds #r #stats #probability #odds #oddsratio

www.montana.edu/rotella/docu...

How would you explain to an everyday person the difference between “you’re have an increased odds” vs. “you’re at an increased risk”?

OR is much easier to model than RR for non-rare outcomes - but balancing the need for interpratbility.
#episky #statsky #oddsratio

The association between birth cohort and marriage hazards is more negative for Black people than White people in odds ratios (panel a) but less negative in average marginal effects (panel b)

#statstab #24 Non-linear model are evil [my title]

Thoughts: When ppl say "just use exp() to get the value on the original scale" a vein pops. NLMs are complicated things. I only tolerate probit model (for SDT reasons).

#oddsratio #riskratio #effects

read.dukeupress.edu/demography/a...