Bioequivalence and Bioavailability Forum

❝ my basic question is why Type three sum of square used in SAS out put in

❝ BE study,

❝ and what is its impotrance and how they calculate?

❝ and how interprete that one?

Hi Shri,

when you "fit a model" you find the set of model constants that give you the smallest possible sums of squares (SS). The fit does not provide info on how much of the SS is associated with each of your factors.
The type III way of finding the SS figures for each factor, is to first fit a model with for example Subject, Sequence and Period in it. Next another model with Subject, Sequence, Period and Treatment is fitted. The difference between the resulting SS in the two models is the SS that is then assigned to the Treatment effect. And so forth for the other factors.

Whether or not this provides a relevant test for your hypothessi I cannot tell - someone with better understanding of statistics should answer this. A good thing about type III is that your ANOVA p-levels will correspond exactly to the P-levels you can derive from ordinary t-tests of the same factors (using e.g. the t-test described by Chow & Liu). SAS invented the type III paradigm and there is VERY MUCH disagreement among statisticians about its relevance.

EM.

dear shri !

the type III sum of squares is a historical term refering to specific hypotheses on available factors included in a statistical model. please refer to this post for more detail link.

hope this helps

martin