Evidence Based Medicine pills
Evidence Based Medicine pills
When multiple hypothesis tests are performed, the probability of committing a type 1 error increases, increasing the risk of detecting false positive effects.
The false discovery rate allows to limit the probability of type 1 error when the number of contrasts is very high, also allowing to control the risk of making type 2 errors and failing to detect true positives.
Cochran's Q is a widely used measure to detect heterogeneity between primary studies in a meta-analysis. Its statistical properties and hypothesis testing are reviewed. Finally, other measures calculated from this value are described, such as the I2 statistic and the H2 statistic, frequently used to quantify the intensity of heterogeneity between studies.
With small sample size, the variance of the Pearson’s coefficient increases, decreasing the precision of its estimates. The use of Fisher's z helps to stabilize the variance and obtain more precise estimates in meta-analyses whose outcome measure is a correlation, and when the sample of primary studies is small.
Multiple regression regularization (shrinkage) techniques can be very useful to address collinearity or overfitting problems. In addition, they can be used to select the independent variables and reduce multidimensionality, achieving more robust and easy-to-interpret models. Ridge, lasso and elastic network regression techniques are described.
The parameters that report on the quality of a multiple logistic regression model and that are usually provided by the statistical programs with which they are carried out are reviewed. Emphasis is placed on the goodness of fit, the predictive capacity and the statistical significance of the model.