Ders 11: Bilimsel tekrarlanabilirlik

MBG1032 Biyoistatistik - Doç.Dr.Alper YILMAZ - 23 Mayıs 2025

ilk video

Dersiminizin ilk haftalarında izlediğimiz video

https://www.youtube.com/watch?v=42QuXLucH3Q

özet

• p-hacking
• dark chocolate case
• 6th sense case
• same data different conclusions case (do black players get more red cards? 29 groups work on same data; 8 groups said “equally likely”, 19 groups said “more likely” 2 groups said “3 times more likely”
• 53 landmark cancer studies only 6 of them can be reproduced

nearly 1/3 of the published results will be wrong even if the things are working normally (ie. statistical power of 80%)

typical statistical power in studies is around 20-40%

Kaynaklar:

• Why Most Published Research Findings Are False PLoS article
• Understanding the p-value - Statistics Help
• Science Isn’t Broken
• I Fooled Millions Into Thinking Chocolate Helps Weight Loss. Here’s How

ikinci video

The Four Most Dangerous Words? A New Study Shows | Laura Arnold | TEDxPennsylvaniaAvenue

özet

• “power posing” çalışması
• beslenme ve sağlık alanında örnek sayısı düşük, yanlış metodoloji ve seçici raporlama
• “file drawer effect” ve “p-hacking”

beslenme üzerine çalışma

başka bir çalışma