: p-value in statistics
In: 10
ELI10.
You want to check if Superman is better than batman at winning the bad guys, so you read one hundred comics of each and take note of how often they win. You note that superman wins 75% of the time and batman 70%. However that’s not a big difference and you haven’t read every single comic and watched every single film, perhaps if you read another one hundred comics, the results would be different. So we need to check how sure we can be that it’s not due to chance.
You need to run some kind of test to know that. First you make a statement that you’re trying to reject (superman and batman are equally as strong). This is your null hypothesis and, for the moment, we will assume it is true because we don’t know any better. Then you make a statement that contradicts the null hypothesis and is what you’re trying to prove (superman is stronger). This is your alternative hypothesis and it’s called alternative because, as I said, we assume the null hypothesis is true until we know better. Thus, you run a statistical test (in this case a T test, but that’s not that relevant right now). This test gives you a p-value which is the odds of getting such a result or a more extreme one of the null hypothesis is true. In our case, it tells you the odds of Superman beating the bad guys at least 5 points more often than batman assuming that they are both as likely to do so. So, if your p-value is 0.04, it means that there’s a 4 out of 100 odds. Very unlikely so the alternative hypothesis is way more likely to be true. If the p-value is larger than a certain random threshold (usually between 0.01 and 0.1) you decide that the odds are big enough for the null hypothesis to stand a good chance of being true and you “fail to reject the null hypothesis” which means “maybe the alternative hypothesis is true but I can’t tell for sure”.
I hope that helped.
Latest Answers