Some statistical markers that talk more about your data
In light of the limitations of p-values explained in the previous lesson, Let’s explore other options that demonstrate reliable statistical analysis.
- Boost your confidence with confidence intervals!
The confidence interval shows the range within which the actual value of a parameter is likely to fall. Unlike p-values, they provide a clearer picture of the estimate range and remove the uncertainty associated with parameter variability. When exploring the statistical difference between two data groups, the confidence interval shows where the mean difference between the two sample groups should lie.
- Measure true impact with an effect size
The effect size determines the practical significance of an effect. It also shows the weight of the difference in a parameter. This gives more meaning to results by including the magnitude of the effect being measured, while the p-value only reflects the statistical probability of significance. Cohen’s D is one standardised method for calculating effect size. It usually lies within a range of 0 to positive infinity. A larger effect size typically implies a more substantial real-world impact.
- Gather evidence with the T-statistic
With the different forms of t-distribution tests, the t-statistic measures how many sample errors the alternative hypothesis is from the null hypothesis. A higher magnitude of the t-statistic suggests more compelling evidence against the null hypothesis. It indicates that the difference between sample means is less likely to be caused by random chance, resulting in rejecting the null hypothesis in support of the alternative hypothesis.
- Too many p-values with no meaningful inference? F-statistics is your guy
F-statistics is particularly useful in statistical tools like ANOVA and regression, which assess multiple variables simultaneously. It gives an overall view of the entire model and provides information beyond what p-values for individual predictors can provide.
Click through the slides below to remind yourself of the other options that demonstrate reliable statistical analysis.