Unlike design-based stereology, model-based stereology experiments are not set up to be assumption free. Instead there is an attempt to change the data after-the-fact to make up for the biased nature of the experiment. An example is the Abercrombie method of correcting cell counts. Instead of setting up the experiment to follow rules so that one and only one cell will be counted at a time (see Number probes), you go ahead and count cell pieces and then massage or correct the data with a formula. You have to know certain information to set up this model, for instance the size of the particles in the z-dimension.
Why not use model based stereology?
Model based stereological methods use approximation methods to describe objects which are being studied. These approximations work only as well as the models truly represent the actual objects. The problems with models can be avoided by simply avoiding the use of models. A common example is, if you want to estimate the number of cells in a given region, rather than counting cell pieces and trying to fix the counts with the Abercrombie method, use thick sections and the optical fractionator probe.
Sponsored by MBF Bioscience
developers of Stereo Investigator, the world’s most cited stereology system