𝗗𝗼 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀 𝗺𝗶𝘀𝗿𝗲𝗮𝗱 𝗽-𝘃𝗮𝗹𝘂𝗲𝘀 𝗹𝗶𝗸𝗲 𝘄𝗲 𝗱𝗼?
McShane, Gal & Duhachek (2025) tested ChatGPT, Gemini, and Claude: p = 0.049 and p = 0.051 often lead to opposite conclusions.
Explained page by page:
▶️ youtu.be/yEZ6IXZ8NzM
#Statistics #PValue #AI #LLM #ResearchMethods #PaperExplained
McShane, Gal & Duhachek (2025) tested ChatGPT, Gemini, and Claude: p = 0.049 and p = 0.051 often lead to opposite conclusions.
Explained page by page:
▶️ youtu.be/yEZ6IXZ8NzM
#Statistics #PValue #AI #LLM #ResearchMethods #PaperExplained
October 19, 2026 at 6:46 PM
𝗗𝗼 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀 𝗺𝗶𝘀𝗿𝗲𝗮𝗱 𝗽-𝘃𝗮𝗹𝘂𝗲𝘀 𝗹𝗶𝗸𝗲 𝘄𝗲 𝗱𝗼?
McShane, Gal & Duhachek (2025) tested ChatGPT, Gemini, and Claude: p = 0.049 and p = 0.051 often lead to opposite conclusions.
Explained page by page:
▶️ youtu.be/yEZ6IXZ8NzM
#Statistics #PValue #AI #LLM #ResearchMethods #PaperExplained
McShane, Gal & Duhachek (2025) tested ChatGPT, Gemini, and Claude: p = 0.049 and p = 0.051 often lead to opposite conclusions.
Explained page by page:
▶️ youtu.be/yEZ6IXZ8NzM
#Statistics #PValue #AI #LLM #ResearchMethods #PaperExplained
Difference between Probability value and Significance value
#Statisticsashtag/Statistics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#Statistics #DataScienceshtag/DataScience" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataScience #HypothesisTesting #PValue #SignificanceLevel #AlphaLevel #DataAnalyticstag/DataAnalytics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataAnalytics #ITFeaturehashtag/ITFeature" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#ITFeature #Statistics #DataScience #DataAnalytics #ResearchMethodology #ITFeature
#Statisticsashtag/Statistics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#Statistics #DataScienceshtag/DataScience" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataScience #HypothesisTesting #PValue #SignificanceLevel #AlphaLevel #DataAnalyticstag/DataAnalytics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataAnalytics #ITFeaturehashtag/ITFeature" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#ITFeature #Statistics #DataScience #DataAnalytics #ResearchMethodology #ITFeature
Difference between Probability and Significance Value
In statistical hypothesis testing, the probability and significance value ($\alpha$) work together to help decide whether experimental findings reflect a real effect or just random chance. Basically, in hypothesis testing, the goal is to see if the probability value (p-value) is less than or equal to the significance level (i.e., is $p \le \alpha$). Key Differences between Probability and Significance Value Practical Example of Probability and Significance Value Pre-Experiment: Setting the Significance Value Post-Experiment: Calculating the p-value Making the Decision Exam pass rate Example At a fundamental level, the difference between the probability value and the significance value is:
itfeature.com
October 17, 2026 at 7:00 PM
Difference between Probability value and Significance value
#Statisticsashtag/Statistics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#Statistics #DataScienceshtag/DataScience" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataScience #HypothesisTesting #PValue #SignificanceLevel #AlphaLevel #DataAnalyticstag/DataAnalytics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataAnalytics #ITFeaturehashtag/ITFeature" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#ITFeature #Statistics #DataScience #DataAnalytics #ResearchMethodology #ITFeature
#Statisticsashtag/Statistics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#Statistics #DataScienceshtag/DataScience" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataScience #HypothesisTesting #PValue #SignificanceLevel #AlphaLevel #DataAnalyticstag/DataAnalytics" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#DataAnalytics #ITFeaturehashtag/ITFeature" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#ITFeature #Statistics #DataScience #DataAnalytics #ResearchMethodology #ITFeature
𝗪𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗮 𝗽-𝘃𝗮𝗹𝘂𝗲 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘁𝗲𝗹𝗹 𝘂𝘀 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗶𝘁 𝗻𝗼𝘁?
In this episode of Paper Explained, I explore common misconceptions about statistical significance, confidence intervals, and how to interpret scientific results.
🎬 www.youtube.com/watch?v=6KV2...
#Statistics #PValue #PaperExplained
In this episode of Paper Explained, I explore common misconceptions about statistical significance, confidence intervals, and how to interpret scientific results.
🎬 www.youtube.com/watch?v=6KV2...
#Statistics #PValue #PaperExplained
P-values: 25 Common Misinterpretations | Paper Explained
YouTube video by Andrea Perlato | Research
www.youtube.com
October 12, 2026 at 9:24 AM
𝗪𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗮 𝗽-𝘃𝗮𝗹𝘂𝗲 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘁𝗲𝗹𝗹 𝘂𝘀 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗶𝘁 𝗻𝗼𝘁?
In this episode of Paper Explained, I explore common misconceptions about statistical significance, confidence intervals, and how to interpret scientific results.
🎬 www.youtube.com/watch?v=6KV2...
#Statistics #PValue #PaperExplained
In this episode of Paper Explained, I explore common misconceptions about statistical significance, confidence intervals, and how to interpret scientific results.
🎬 www.youtube.com/watch?v=6KV2...
#Statistics #PValue #PaperExplained
what I am missing from the p.value discourse are applied examples of when a p.value less then or greater than .05 is approporiate. Eg, given the questions outcomes / applications/ sensitivities/ use cases, how they determined what is the more appropriate pvalue for that test would help alot
September 20, 2026 at 8:24 PM
what I am missing from the p.value discourse are applied examples of when a p.value less then or greater than .05 is approporiate. Eg, given the questions outcomes / applications/ sensitivities/ use cases, how they determined what is the more appropriate pvalue for that test would help alot
I want one with a low pvalue
August 16, 2026 at 8:35 PM
I want one with a low pvalue
День начался с очередного вопроса от очередного менеджера, не виновато ли pvalue в том, что эксперимент серый.
Даже не знаю, что отвечать на такое, хоть доку заготовить и кидать. Но вроде для продакта, постоянно крутящего экспы, это базовые знания.
Даже не знаю, что отвечать на такое, хоть доку заготовить и кидать. Но вроде для продакта, постоянно крутящего экспы, это базовые знания.
February 12, 2025 at 6:25 AM
День начался с очередного вопроса от очередного менеджера, не виновато ли pvalue в том, что эксперимент серый.
Даже не знаю, что отвечать на такое, хоть доку заготовить и кидать. Но вроде для продакта, постоянно крутящего экспы, это базовые знания.
Даже не знаю, что отвечать на такое, хоть доку заготовить и кидать. Но вроде для продакта, постоянно крутящего экспы, это базовые знания.
Вот и настал момент на работе когда пригодилось знание определения pvalue, которое так любят обсуждать на собесах.
Разработчик спросил, что это за чиселка в табличке.
Разработчик спросил, что это за чиселка в табличке.
January 9, 2025 at 12:05 PM
Вот и настал момент на работе когда пригодилось знание определения pvalue, которое так любят обсуждать на собесах.
Разработчик спросил, что это за чиселка в табличке.
Разработчик спросил, что это за чиселка в табличке.
Yeah I’m just not seeing the added value of recalculating a pvalue to apply a subjective heuristic for gauging the plausibility of an effect and it seems quite pitfall ridden. Prior predictive simulation or some such would make bit more sense.
July 22, 2025 at 11:58 PM
Yeah I’m just not seeing the added value of recalculating a pvalue to apply a subjective heuristic for gauging the plausibility of an effect and it seems quite pitfall ridden. Prior predictive simulation or some such would make bit more sense.
Are biologists 'overdepending' on significance testing? Our Tom Crowther in @PNASNews Early Edition http://m.pnas.org/content/early/2015/08/25/1513283112.short?rss=1).%25C2%25A0 #pvalue
December 4, 2024 at 2:12 PM
Are biologists 'overdepending' on significance testing? Our Tom Crowther in @PNASNews Early Edition http://m.pnas.org/content/early/2015/08/25/1513283112.short?rss=1).%25C2%25A0 #pvalue
The Critical Value and the p-Value Approach to Hypothesis Testing… shorturl.at/ruoIk
#statistics #pvalue #hypothesis #testing #snoopdogg
#statistics #pvalue #hypothesis #testing #snoopdogg
December 18, 2024 at 6:45 PM
The Critical Value and the p-Value Approach to Hypothesis Testing… shorturl.at/ruoIk
#statistics #pvalue #hypothesis #testing #snoopdogg
#statistics #pvalue #hypothesis #testing #snoopdogg
#Ecologists explaining less but with more confidence? http://news.sciencemag.org/biology/2014/08/ecology-explaining-less-and-less #statistics #pvalue
May 24, 2025 at 2:05 AM
#Ecologists explaining less but with more confidence? http://news.sciencemag.org/biology/2014/08/ecology-explaining-less-and-less #statistics #pvalue
i have access to instagram and reddit again and oooooooo boy. let’s just say the regression coefficient of my mental health vs access to those apps has a pvalue<0.05.
February 5, 2025 at 11:11 PM
i have access to instagram and reddit again and oooooooo boy. let’s just say the regression coefficient of my mental health vs access to those apps has a pvalue<0.05.
Ended up using this one for CIs. Followed the link to Kristoffer Magnusson's website and found one that works well enough for CLT (rpsychologist.com/pvalue/)
November 1, 2023 at 1:43 PM
Ended up using this one for CIs. Followed the link to Kristoffer Magnusson's website and found one that works well enough for CLT (rpsychologist.com/pvalue/)
Suppose you have two groups of very large size. You are curious how easily a difference D in mean height btw groups could arise by chance if they’re from the same population.
Then z ~ (D / sigma_hat) and Pr(Z > z | Z ~ N(0, 1)) is your pvalue for this test of a difference in means. Z is approximate
Then z ~ (D / sigma_hat) and Pr(Z > z | Z ~ N(0, 1)) is your pvalue for this test of a difference in means. Z is approximate
January 30, 2024 at 4:27 AM
Suppose you have two groups of very large size. You are curious how easily a difference D in mean height btw groups could arise by chance if they’re from the same population.
Then z ~ (D / sigma_hat) and Pr(Z > z | Z ~ N(0, 1)) is your pvalue for this test of a difference in means. Z is approximate
Then z ~ (D / sigma_hat) and Pr(Z > z | Z ~ N(0, 1)) is your pvalue for this test of a difference in means. Z is approximate
This conflates the likelihood of the null (the joint probability of observing the data given the null model) with a p value. These are not the same.
Frequentist stats has many problems. The tendency of most people to view a pvalue as a point estimate rather than a tail probability is one if them.
Frequentist stats has many problems. The tendency of most people to view a pvalue as a point estimate rather than a tail probability is one if them.
January 30, 2024 at 4:02 AM
This conflates the likelihood of the null (the joint probability of observing the data given the null model) with a p value. These are not the same.
Frequentist stats has many problems. The tendency of most people to view a pvalue as a point estimate rather than a tail probability is one if them.
Frequentist stats has many problems. The tendency of most people to view a pvalue as a point estimate rather than a tail probability is one if them.
Prashant Kishor’s Jan Suraaj leads in early Bihar trends on three seats
November 14, 2025 (Punjab Khabarnama Bureau) : Prashant Kishor's Jan Suraaj Party (JSP) was leading in at least four assembly constituencies in Bihar in early Bihar election counting trends, provided by Pvalue Analytics. The…
November 14, 2025 (Punjab Khabarnama Bureau) : Prashant Kishor's Jan Suraaj Party (JSP) was leading in at least four assembly constituencies in Bihar in early Bihar election counting trends, provided by Pvalue Analytics. The…
Prashant Kishor’s Jan Suraaj leads in early Bihar trends on three seats
November 14, 2025 (Punjab Khabarnama Bureau) : Prashant Kishor's Jan Suraaj Party (JSP) was leading in at least four assembly constituencies in Bihar in early Bihar election counting trends, provided by Pvalue Analytics. The trends defied most exit polls, which projected that the debutant JSP is expected to draw blank in the crucial elections. Follow Bihar election results live updates…
punjabkhabarnama.com
November 14, 2025 at 4:12 AM
Prashant Kishor’s Jan Suraaj leads in early Bihar trends on three seats
November 14, 2025 (Punjab Khabarnama Bureau) : Prashant Kishor's Jan Suraaj Party (JSP) was leading in at least four assembly constituencies in Bihar in early Bihar election counting trends, provided by Pvalue Analytics. The…
November 14, 2025 (Punjab Khabarnama Bureau) : Prashant Kishor's Jan Suraaj Party (JSP) was leading in at least four assembly constituencies in Bihar in early Bihar election counting trends, provided by Pvalue Analytics. The…
P values @naturemethods "need to bring many sources of evidence (...) before drawing scientific conclusions#pvalue
go.nature.com/2nOFiGQ
go.nature.com/2nOFiGQ
Interpreting P values
Nature Methods - A P value measures a sample's compatibility with a hypothesis, not the truth of the hypothesis.
go.nature.com
January 20, 2025 at 7:08 PM
P values @naturemethods "need to bring many sources of evidence (...) before drawing scientific conclusions#pvalue
go.nature.com/2nOFiGQ
go.nature.com/2nOFiGQ
Z-Skoru ve P-Değerlerine Bağlı Anlamlılık Bölgeleri
#ZSkoru #PDeğeri #ZScore #PValue #SıfırHipotezi #NullHypothesis
#ZSkoru #PDeğeri #ZScore #PValue #SıfırHipotezi #NullHypothesis
March 17, 2025 at 10:54 PM
Z-Skoru ve P-Değerlerine Bağlı Anlamlılık Bölgeleri
#ZSkoru #PDeğeri #ZScore #PValue #SıfırHipotezi #NullHypothesis
#ZSkoru #PDeğeri #ZScore #PValue #SıfırHipotezi #NullHypothesis
Obviously the man is a fan of the Pvalue
a man wearing a sombrero and a red poncho is holding maracas .
ALT: a man wearing a sombrero and a red poncho is holding maracas .
media.tenor.com
April 18, 2025 at 12:22 AM
Obviously the man is a fan of the Pvalue
Eg: "using the best pvalue from preliminary measurements n=50 it is clear that the n=200 is well powered across all 1000 metabolites even with bonferroni correction".
Don't make me, the reviewer, do your power calculations.
Don't make me, the reviewer, do your power calculations.
May 10, 2026 at 9:26 AM
Eg: "using the best pvalue from preliminary measurements n=50 it is clear that the n=200 is well powered across all 1000 metabolites even with bonferroni correction".
Don't make me, the reviewer, do your power calculations.
Don't make me, the reviewer, do your power calculations.
The consortium, led by the statistician Peter Donnelly settled on a raw Pvalue of 5e-8 or less, though making more of a Bayesian argument of priors and density of true findings and less of a frequentist number of tests (despite it being a frequentist like probability under the null)
December 27, 2024 at 5:53 PM
The consortium, led by the statistician Peter Donnelly settled on a raw Pvalue of 5e-8 or less, though making more of a Bayesian argument of priors and density of true findings and less of a frequentist number of tests (despite it being a frequentist like probability under the null)
Ko wen-je received strong support from the younger generations. How strong is it? There is a high correlation between the percentage of people voting for Ko and people below 40. The scatter plot shows the relationship between two data. The Pearson correlation result is 0.685 and pvalue is 0.0.
January 15, 2024 at 7:05 AM
Ko wen-je received strong support from the younger generations. How strong is it? There is a high correlation between the percentage of people voting for Ko and people below 40. The scatter plot shows the relationship between two data. The Pearson correlation result is 0.685 and pvalue is 0.0.
while thinking about things I am definitely not going to include in this week's homework, I accidentally discovered that works on AO3 with foot fetish content attract significantly (MannWhitney-U=324761.5, pvalue=0.025) more kudos.
I am going outside. I may be quite some time.
I am going outside. I may be quite some time.
November 19, 2024 at 2:34 PM
while thinking about things I am definitely not going to include in this week's homework, I accidentally discovered that works on AO3 with foot fetish content attract significantly (MannWhitney-U=324761.5, pvalue=0.025) more kudos.
I am going outside. I may be quite some time.
I am going outside. I may be quite some time.
It's a little odd. I am plotting the log10 of the -log10(pvalue) on the x axis so in this plot the null signal drops off much faster than what we observe in the @gwascatalog.bsky.social
I'm curious if anyone has a hypothesis why the lead SNP p-values would have this distribution.
I'm curious if anyone has a hypothesis why the lead SNP p-values would have this distribution.
September 28, 2025 at 1:50 PM
It's a little odd. I am plotting the log10 of the -log10(pvalue) on the x axis so in this plot the null signal drops off much faster than what we observe in the @gwascatalog.bsky.social
I'm curious if anyone has a hypothesis why the lead SNP p-values would have this distribution.
I'm curious if anyone has a hypothesis why the lead SNP p-values would have this distribution.
Aargh. The curse of the underflow p value strikes the MVP study.
In picking the lead snps the authors (inadvertantly) picked the first SNP with an underflow pvalue. This can be up to 500kb from the true lead snp.
All the strong pvalues listed from verma et al in @gwascatalog.bsky.social are suspect
In picking the lead snps the authors (inadvertantly) picked the first SNP with an underflow pvalue. This can be up to 500kb from the true lead snp.
All the strong pvalues listed from verma et al in @gwascatalog.bsky.social are suspect
See now this is what I'm talking about. A massive study (MVP) with 80,000 GWAS associations and the 1,000 most significant are corrupted because the authors didn't provide their actual p-values.
www.ebi.ac.uk/gwas/publica...
Don't do this
www.ebi.ac.uk/gwas/publica...
Don't do this
February 11, 2025 at 2:52 PM
Aargh. The curse of the underflow p value strikes the MVP study.
In picking the lead snps the authors (inadvertantly) picked the first SNP with an underflow pvalue. This can be up to 500kb from the true lead snp.
All the strong pvalues listed from verma et al in @gwascatalog.bsky.social are suspect
In picking the lead snps the authors (inadvertantly) picked the first SNP with an underflow pvalue. This can be up to 500kb from the true lead snp.
All the strong pvalues listed from verma et al in @gwascatalog.bsky.social are suspect