AJ's statistical analysis is a good way to decide how many boxes to count. I also look at box to box standard deviation and make sure it is reasonable.
The average number of cells in a square is N*p where N is the number of cells you pipeted and p is the probability that a cell winds up in the square you are looking at which is the the ratio of the volume of the square to the volume you pipet. If you have a sample with 10,000,000 cells/mL, a large square has volume (which it typically does) of 0.0001 mL and you pipet 0.1 mL then N = 1,000,000 cells; p = 0.0001/0.1 = 0.001 and you expect to see 1,000,000 * 0.001 = 1000 cells. The standard deviation in the number of cells in a square is sqrt(N*p*(1-p)) which, as p is 0.001, is close to sqrt(N*p). Thus, as long as the amount you pipet is larger relative to a square size the standard deviation is closely approximate to the square root of the average cell count. For the example of 10,000,000 cells/mL with an average large square count of 1000 the standard deviation between large squares would be sqrt(1000) = 31.6.
The same applies to the small squares except that p is even smaller and the approximation even better. My hemocytometer has 16 small squares per large (many have 25) so I'd expect 1000/16 = 62.5 cells per small square and the standard deviation in small square count should thus be sqrt(62.5) = 7.9.
..keep in mind that dilutions and homogenization are also critical.
It should be made very clear that the uncertainty of 1/sqrt(cells_counted) uncertainty I mentioned in the earlier post is in addition to other uncertainties. Each time you dilute you introduce uncertainty not just from the uncertainties in the measurements of the volumes (though those are the major factor no doubt) but because you are making a random draw from a larger population of suspended yeast cells. There is a cell density in the fermentor - that's what we are trying to determine. The cell density in a hydrometer sample, for example, is not the same. Its expected cell density is but its actual cell density is a random variable with standard deviation dependent on the ratio of the sample size to the full volume. The smaller the sample the larger that deviation.
One thing that I have found recently is that a pipette can retain 5% of the previous sample clinging to it's walls. So if you wash your pipette and then pull a sample it could be 5% diluted.
There are two types of pipet - blow down and non blow down. In a blow-down pipet you squeeze the pipetter bulb to force air through the pipet in order to expel all (or most all) of any liquid retained. Pipets are marked with either a double band or single band at the top to tell you which type you have.
