P-values are a way to measure whether the results of an experiment or study are likely due to chance or if they reflect a real effect (intervention). They are a statistical tool to help us decide if what we're seeing is meaningful. Remember a P value mathmatically states whether there is a real effect but this might not mean its felt by our patients.
Test hypotheses: They help determine if a null hypothesis (the idea that there is no effect or no difference) can be rejected.
Make decisions: By looking at the p-value, we assess the strength of the evidence against the null hypothesis.
First learn to convert the P Value to a percentage (below). Then place this in a sentance. A percentage greater than 5% is considered not significant (due to chance)
0.06 = 6%
There is a 6% probability that the effect is by chance, therefore the real effect is insignificant
0.1 = 10%
There is a 10% probability that the effect is by chance, therefore the real effect is insignificant
0.01 = 1%
There is a 1% probability that the effect is by chance, therefore the real effect is significant
0.53 = 53%
There is a 6% probability that the effect is by chance, therefore the real effect is insignificant
A small p-value (typically < 0.05) is significant and suggests strong evidence against the null hypothesis, so you might reject it.
A large p-value (typically > 0.05) is insignificant and suggests weak evidence against the null hypothesis, so you might not reject it.
P-values were developed in the early 20th century by Ronald A. Fisher, a statistician. He introduced the concept as a way to measure evidence in favor of or against a hypothesis. Fisher's work laid the foundation for modern hypothesis testing.
£65k
£50k
£30k
£25k
£20k
£245 Billion (Elon Musk)
£50k
£30k
£25k
£20k
Data is skewed (e.g., house prices with extreme high values).
Outliers are present, as the median is not affected by extreme values.
Ordinal data needs summarizing (e.g., survey rankings).
The dataset is small or uneven, making the median more representative.
Summarising discrete data like counts (e.g., hospital visits).
Symmetrical and not skewed.
Without extreme outliers.
Numerical and continuous (e.g., heights, weights, or test scores).