Obviously, 0.48 is less than 0.05, so we reject the null hypothesis. After successfully conducting our experiment and recording our observations, we compute a P-value of 0.48, meaning that we have about a 48% chance of obtaining data at the same level of statistical significance as our sample data under the assumption that Coca-Cola and physique are independent of each other. We will test whether Group A has a generally lower heart rate than Group B after a short cardiovascular exercise. We take a sample size of twenty people that drink Coca-Cola every day (Group A) and another twenty people that abstain from America’s beloved sugary drinks (Group B).

Now that we set the scene, let us commit p-hacking. We are researchers paid by soda distributors, and our job is to unethically conclude that these carbonated drinks statistically improve physical health.

Let’s Do Some Data Dredging: P-Hacking With Our Coca-Cola Example The p-value we compute must be less than our chosen benchmark value for us to conclude statistical significance. If you think about it, the smaller our computed p-value is, the more unlikely it is that the data we obtained was by pure coincidence. The most commonly selected benchmark value for p-value analysis is 0.05, or a 5% random chance that we would have received our test results in a universe where daily Coca-Cola consumption has no significant positive correlation with fitness. The statistical significance of data is communicated through something called a p-value, which is the probability of obtaining our results in a universe where our null hypothesis is true. P-hacking can lead to academic papers headlined with false positives, such as tobacco smoking improving health or vaccinations damaging the human body. If a researcher runs ninety-nine statistically insignificant experiments before obtaining a statistically significant one, and they only report the significant one in a scientific paper, that individual is guilty of p-hacking. Thus, in our Coca-Cola experiment, our null hypothesis would be the claim that everyday consumption of Coca-Cola has no significant positive correlation with fitness.ĭata dredging, also called p-value hacking, stems from reporting cherry-picked statistically significant results from a set of tests while intentionally leaving out the necessary context of the majority statistically insignificant ones. In other words, a null hypothesis is a statement researchers build just to hopefully knock down. In hypothesis testing, researchers formulate a null hypothesis - which is the idea that the variables we are testing do not affect the results.

Data dredging/ p-hacking