I am looking at the distribution of lamps across time (N=687). Most of my lamps (cases) are coded into 50-year brackets within a single categorical variable (0-50 CE =1, 50-100 CE =2, 100-150 CE =3, 150-200 CE =4 and so forth). I will be conducting chi-square analyses or Fishers exact test, where the cell totals are below 5.

Unfortunately, a number of the lamps cannot be neatly dated. For instance, I have a lamp that is dated 50-150 CE, which covers a 100-year period, as opposed to 50-years. There are about 50 of these problem lamps in total.

Some have argued that I needs= to either force these lamps into a 50-year bracket and make some justification for this in the body of the thesis, or exclude them from the analysis. However, one of my supervisors has insisted that these lamps should be counted twice, I.e. coded as both 2 and 3.

Obviously the problem with this approach is that a 'dummy' case would need to be created for each one of the problem lamps, resulting in an increasing N. This is a problem due to the number of problem lamps; the N will jump by at least 50.

My question is this: Is there a way to enter a single case (lamp) twice without increasing the total number of the cases? Or, alternatively, is there another way you can think of dealing with this issue?

Thanks in advance.