|Title||»||Prediction versus confidence intervals in peak area plots|
|Assigned To||»||Brendan MacLean|
Currently in our Group By replicate annotation peak area plots, Skyline shows whiskers for + or - in standard deviation. In performing this type of grouping and investigator is usually interested in some expression of overall variance. So, what we have is nice.
But, if they are interested in whether two groups might yield a t-test for significantly different mean values, this plot doesn't really tell them much. What they need instead is an error bar based on the standard error. It might also be interesting to allow the option of specifying a level of confidence (95% or 99% are common), and show these standard error-based CIs. Similarly, if they are interested in how strong a predictor the value might make, then the current standard deviation-based whiskers are more informative, but we might still be able to improve by showing a prediction interval at a user specified percentage confidence. For instance, if the 95% prediction intervals do not overlap, then the peptide is a very strong predictor all by itself.