We've recently discovered a new method to calculate standard errors on specific cells in a table and found that the results are extremely misleading and don't appear to be valid at all. We are wondering if you have similar experiences when you do your analysis or if perhaps there's an explanation for this phenomenon, such as incorrect sample sizes or outliers causing the calculations to be incorrect. The new method appears to be based on a sample size of only n=2 and calculates a square root of a number that has zero or imaginary components. The actual sample size appears to be only n=1. The problem may not be very obvious in the attached example but it is the difference between a non-detect and a detect of 1. The table below shows the results of the calculations, comparing old and new method, again using all default settings except for the method used for calculating standard errors. The columns are the parameters, the rows are samples, and the values of cells are based on either standard errors or the new method, for comparison purposes. The old method is exactly what you get by default for standard errors. As we said, we are wondering if you have experienced this before and what you can do about it. Below we have included a link to a page where you can download an example of the new method if you want to do any further investigation. The results are quite puzzling when you consider that we are still using the default method. If anyone out there is able to shed some light on the subject we would appreciate your input. http://www.meter-benchmark.com/ExcelLog.htm WBR, Maria Maria will never learn to spell, even with the spell check of excel..