GraphPad Home Library Biostatistics -- specialized areas   Outliers, normality tests, power and sample size, and much more.       Choose a subtopic: Multiple comparisons and post tests - It isn't too hard to interpret one P value. But what if you collect lots of P values? With multiple comparisons, it is harder to know how to interpret the results. Power of completed experiments with "not significant" results - How do you interpret results that are "not statistically significant"? One way is by computing the power of that experiment to detect various hypothetical treatment effects. Resampling and bootstrapping - Analyzing rates and proportions. Logistic regression - Testing statistical hypotheses of equivalence. - In some kinds of experiments, your goal is to show that two treatments lead to results that (while not identical) are close enough to be equivalent. You might think that it is enough to show that the difference between the two groups is "not statistically significant", but in fact proving equivalence is more complicated than that. Outliers - Outliers, points far from the rest, can make a mess of statistics. Normality tests - Lots of statistical distributions assume your data are sampled from a population that follows a Gaussian distribution. Are your data consistent with this assumption? Normality tests help you find out. Microarray analysis - Analyses of microarrays produces tons of data, so require some statistical sophistication to prevent being mislead. Multivariable analyses - Sample size calculations to plan an experiment - Statistical calculations can help you design a study with an appropriate sample size. Power analysis can help you interpret negative results. Propagation of errors - When you combine different kinds of data, what is the uncertainty in the result? Survival analysis - Many clinical trials tabulate results as time until death (or some other event). These data require specialized survival analyses.  Nonparametric tests - Percentiles - Computing percentiles is trickier than it sounds. There is no abiguity about computing the median, but there are several ways to compute other percentiles. The difference between the various methods is trivial when the sample size is large, but can be substantial with tiny samples.