Basic statistics: A research primer for low- and middle-income countries

Statistics can be used to describe data or make inferences about populations using samples. Median values (the 50th percentile) better represent central tendency of data samples than means (averages), particularly when data have extreme values. Errors resulting from use of inferential statistics when using classical hypothesis testing include type I (finding a difference between groups when one does not exist) and type II (failure to find a true difference) errors. Confounding variables (those that vary with both the dependent variable and independent variable) may lead to spurious associations. Classical hypothesis testing and reporting only p-values tends to be greatly overused and overemphasized. Confidence intervals provide a range of values for a sample within a certain probability (commonly 95%). Confidence intervals can thus describe sizes of likely differences between samples, and are much more clinically useful information than only p-values. Before doing a study, the required sample size should be calculated to assess study feasibility. Doing so requires specification of the acceptable risk of type I and II errors and the size of the lowest clinically meaningful difference between groups.