This semester, I have been working on a project as part of an independent course at the college level. In a couple of weeks, I will have to present my results at the NEERS conference in ME and at the MAA conference in CA. I have been working with Battelle (Battelle) in viewing the statistical methods currently employed to determine whether an estuary should remain open or shut. Based on fecal coliform levels and an exceedance level mandated by the EPA, an estuary remains open or closed whether it's above or below this exceedance level.
My presentation includes showing confidence intervals for a lognormal distribution, using a 95% confidence level, to try give a better idea of whether these estuaries should remain open or should be closed. In particular, I have confidence intervals for the geomean and the 90th percentile. Each estuary has multiple stations, and the confidence intervals show range of fecal coliform within each station.
Obviously part of my presentation needs to discuss the idea behind confidence intervals, showing that the data is assumed to be drawn from a lognormally distributed population, giving a brief comparison of a sample versus the population, showing what the geomean is and how it shifts back to the normal scale from the lognormal scale, showing that the 90th percentile on the lognormal scale is the 90th percentile on the normal scale, and so on and so forth.
I was wondering if there was any easy way of creating two scales and showing how everything transforms itself. (Note, this will all be done in a powerpoint presentaiton). That is, we start by taking the log of all the data, taking the arithmetic mean, exponentiate that for the geomean, maybe explain a non-centrality t, etc.
I'm open to suggestions. Thanks.