#StatsTwitter @gonen_mithat #UninformativeStatistics 3: Can anyone cite a paper where it was of value to show a graph of the ROC curve for a prediction model ? Remember that sensitivity and specific have little clinical meaning without incorporating prevalence.
#StatsTwitter #UninformativeStatistics 2: Can anyone cite a paper where showing the I-squared statistic in a metaanalysis told us more than the p-value for heterogeneity and the forest plot? Remember that I^2 is a hybrid of the degree of variation and the sample size
#StatsTwitter #UninformativeStatistics 1: Can anyone cite a paper where it was of value to show the Hosmer Lemeshow statistic for goodness of fit for a logistic regression? Remember the null hypothesis here is “good fit”.
A planned set of tweets: a call out to anyone (e.g. on #StatsTwitter) to provide examples where a given statistical approach is actually informative. Many papers include #UninformativeStatistics simply because it is common practice, not because it helps understand the data.
