This is the start of the server. It is aesthetically crude (as you can see!) and has only a few fairly simple apps, but it's a start!

Apps

• Simple apps to get useful values from a few inputs
• App that gives you the RCI (Reliable Change Index) for given baseline score SD, reliability and the inclusion interval you want (usually .95, i.e. 95%). Uses the getRCIfromSDandAlpha() function from the CECPfuns R library but the maths is trivial.
• App which extends that to give you a plausible interval around that RCI using the confidence interval for the SD given your n.
• App that gives you the CSC (Clinically Significant Change) method c cut off value for given referential help-seeking and non-help-seeking means and SDs. The maths is trivial for the CSC but this does show the plot of the probabilities underpinning the CSC and that it balances the misclassification rates (given truly Gaussian distributions!).
• Confidence interval apps
• Apps given confidence intervals from observed counts or statistics and dataset sizes These can be useful if you have these counts and/or stats from your own data or if you are looking at reports that give those but don't give confidence intervals.
• App that gives you a confidence interval around an observed proportion given the total n, x the (smaller) number of interest within that n and the width of CI you want (usually .95, i.e. 95%). Uses the binconf() function from the Hmisc R library which uses the Wilson method to get the CI.
• App that gives you a confidence interval around the difference between two independent proportions given the total n, x in each dataset/group, the numbers of interest within each group n and the width of CI you want (usually .95, i.e. 95%). Gives what I call a "biplane plot" of the two CIs.
• App that gives you a confidence interval (CI) around an observed mean given that value, either the observed SD or SE, the width of CI you want (usually .95, i.e. 95%) and the n of the dataset from which the alpha was obtained. The maths is parametric and can be sensitive to non-Gaussian distributions, bootstrap CI definitely preferred if you have the raw data.
• App that gives you a confidence interval (CI) around an observed standard deviation given that value, either the observed SD or SE, the width of CI you want (usually .95, i.e. 95%) and the n of the dataset from which the alpha was obtained. Again, the maths is parametric and can be sensitive to non-Gaussian distributions, bootstrap CI definitely preferred if you have the raw data.
• App that gives you a confidence interval around an observed Pearson correlation given the total n, R the correlation and the width of CI you want (usually .95, i.e. 95%). Uses basic parametric theory so treat the CI as indicative rather than definitive. However, if all you have is the n and correlation, it's the best you can do!
• App that gives you a confidence interval around an observed Spearman correlation given the total n, Rs the correlation and the width of CI you want (usually .95, i.e. 95%). I used to say the Pearson CI could be used but that is a very poor approximation, this is better.
• App that gives you a confidence interval (CI) around an observed Cronbach alpha value given that value, the width of CI you want (usually .95, i.e. 95%) and the n of the dataset from which the alpha was obtained. The maths is parametric but fairly robust to non-Gaussian distributions.
• App that gives you a confidence interval (CI) around observed quantiles of a distribution given the data, the quantiles you want and the width of CI you want (usually .95, i.e. 95%) and the n of the dataset from which the alpha was obtained. Uses Nyblom's method to get robust CIs for the quantiles.
• Apps to give bootstrap based confidence intervals from raw data
Still to come: I need to work out good ways for users to upload raw data into shiny apps first.
• Apps to give plots
• From population parameters or sample statistics
  • App that gives a plot of the CIs around an observed Pearson correlation for a range of dataset sizes (n). The CI can be anything from .7 to .999 but defaults to .95, i.e. 95%. Parametric CI as it assumes you only have an R value of interest to you, not data. Probably mainly for teaching/illustration.
• From raw data
• App that gives you a confidence interval (CI) around observed quantiles of a distribution given the data, the quantiles you want and the width of CI you want (usually .95, i.e. 95%) and the n of the dataset from which the alpha was obtained. Uses Nyblom's method to get robust CIs for the quantiles.
• Didactic and modelling apps
• Model of a simple screening design taking the n, prevalence and the sensitivity and specificity of the screening test. Gives full screening table and the PPV and NPV. Changing the prevalence allows you to see how much this affects PPV and NPV. Created to complement OMbook Glossary entry about screening.

Email update list

There is now an Email announcement list, never updating more than monthly, where I will put up developments of new apps here but also a summary of updates to the online glossary and new posts in the Rblog. You can sign up for the update list here to be alerted to new things.

Last updated 27.viii.23