Work mostly by CE other than CORE (for which, see : www.coresystemtrust.org.uk)
Mostly “cell” in statistical work refers to a group of observations sharing two or more characteristics: it’s a “cell” in a crosstabulation.
So in a simple crosstabulation of clients who came to all sessions offered versus those who DNA’ed (Did Not Attend) some against gender there are four cells. Here’s an utterly implausibly well balanced cross-tabulation of those two variables.
The four “cells” that are of interest statistically are the four in the centre, the other boxes are labels (vital of course!) and totals. The totals of course are determined by the numbers in the four cells so are secondary but it’s often helpful to see them.
By tradition what’s in the top left box in the cross-tabulation tells you what is being said in the main cells, here n, the number of observations with values of the column and row labels
Cell size is simply the number of observation in a cell so here all four cells have size 125. Sometimes statisticians use “cell” and “cell size” for one categories. For instance they might say that the minimum cell size for the gender split is 250.
One reason to be interested in cell size is that precision of estimation of the proportions or of the associations between the two variables, or statistical power to detect unbalanced proportions or systematic associations, is determined by the smallest cells much more than by the total sample size.
AssociationConfidence intervalsPrecisionStatistical powerStatistical testsCategorical/nominal variables
Chapters 5, a bit in 6 and more in 7.
Powered by BetterDocs