Dealing with 'zero counts' in keyword analysis
One problem with keyword analysis is that the target corpus will likely include words that do not occur in the reference corpus. In calculating various measures of keyness, this would result in a division by zero, which is mathematically impossible, as far as I know. The default way of dealing with this is to assign words that do not occur in the reference corpus a frequency of 0.5, but this introduces the risk of a result in which such keywords dominate the top positions, because their keyness is inflated.
To remedy this problem, I have added an option to the Keyword Analysis Tool which let's you choose to either go with the default of assigning a 0.5 frequency to 'zero counts', or to simply discard them from all calculations, resulting in keywords that have a minimal frequency of 1 in the reference corpus.
Dealing with 'zero counts' in the Keyword Analysis Tool
There is no real wrong or right way to do this, but at least now you have a choice. Have fun!