**Biplot based on the original two-way table**

From *Models*, click *Centering, *then *No Centering*.

This biplot displays E+G+GE as well as the Grand mean. The advantage is that the real data is approximated and can be visualized if the biplot displays most of the variation in the data. For example, all the entry-tester combinations are positive as there are no negative values in the original data. The disadvantage is that the grand mean masks other variations so that the variations among testers (E), among entries (G), and particularly, the interactions between the entries and the testers (GE), all appear to be small (note that PC1 explained 99% of the total variation!). This problem becomes more severe as the absolute value of the grand mean becomes greater. This model is useful *only *when the grand mean is close to 0, such as a covariate-effect tableof correlation coefficients.

From the biplot, we can see that testers 5 and 6 had higher mean values whereas testers 9, 3, and 4 had smaller mean values. Also, it is barely obvious that Entries 5 and 7 had higher mean values than other entries (e.g. Entry3). Some interactions can be seen too, although one cannot be sure if this is real, because PC2 explained only 1% of the total variation. For example, entries 1, 2, 5 and 6 seem to have interacted positively with testers 5 and 7.

It can be seen that the pattern in this biplot is very similar to that of model 0, except the "resolution" is not as good. Also, compare biplots based on other models based on the same data.