The kurtosis of a data set provides a measure of the peakedness of the distribution of the data, relative to the normal distribution.
A positive kurtosis value indicates a relatively peaked distribution and a negative kurtosis value indicates a relatively flat distribution.
The kurtosis is described further on the Wikipedia kurtosis page.
The Excel KURT function calculates the kurtosis of a supplied set of values.
The syntax of the function is:
where the number arguments are a minimum of four data values for which you want to calculate the kurtosis.
Each of the number arguments can be an individual numeric value or an array of values. These can be supplied to the Kurt function either:
Note: if logical values and text representations of numbers are typed directly into the Kurt function, they are included in the calculation. However, logical values and any text values (including text representations of numbers) that are stored within an array of cells are ignored.
In Excel 2007 and later versions of Excel, the Kurt function can accept up to 255 number arguments, but in Excel 2003, the function can only accept up to 30 number arguments.
Column A of the above spreadsheet on the right shows an array of data, alongside the associated data distribution chart.
The kurtosis of the data in column A of the spreadsheet can be calculated using the Excel Kurt function as follows:
This gives the result 0.532657874, indicating a distribution that is relatively peaked (compared to the normal distribution).
Further details and examples of the Excel Kurt function are provided on the Microsoft Office website.
If you get an error from the Excel Kurt Function, this is likely to be one of the following:
#DIV/0!    Occurs if either:

#VALUE!    Occurs if any of the supplied number arguments that are supplied directly to the function are not recognised as numeric values. (Note that if the Kurt function is provided with a reference to a range of cells, any text values within this cell range are simply ignored and do not cause the function to return an error). 