This is the same regardless of repeated measures. The df for the systematic differences among rows equals number of rows -1, which is 1 for this example.The df for interaction equals (Number of columns - 1) (Number of rows - 1), so for this example is 2*1=2.It is the same regardless of any assumptions about repeated measures. This is the total number of values (18) minus 1. In the final columns, some of that variation can also be attributed to interaction between subjects and either rows or columns. The SS for residual is smaller when you assume repeated measures, as some of that variation can be attributed to variation among subjects.The SS values for the interaction and for the systematic effects of rows and columns (the top three rows) are the same in all four analyses.This measures the total variation among the 18 values. Now look at the SS columns for the analyses of the same data but with various assumptions about repeated measures. The last row shows the total amount of variation among all 18 values.The second to the last row shows the variation not explained by any of the other rows.The third row show the the amount of variation that is due to systematic differences between the columns.The second row show the the amount of variation that is due to systematic differences between the two rows.Equivalently, it quantifies how much variation is due to the fact that the differences among columns is not the same for both rows. It quantifies how much variation is due to the fact that the differences between rows are not the same for all columns. The first row shows the interaction of rows and columns.I rearranged and renamed a bit so the four can be shown on one table.įocus first on the sum-of-squares (SS) column with no repeated measures: The values below are all reported by Prism. The table below shows the ANOVA tables for the four analyses. The colors are repeated between tables, but this means nothing. Each color within a table represents one subject. The tables below are color coded to explain these designs. I analyzed the data four ways: assuming no repeated measures, assuming repeated measures with matched values stacked, assuming repeated measures with matched values spread across a row, and with repeated measures in both directions. There were no missing values, so 18 values were entered in all. To create the examples below, I entered data with two rows, three columns, and three side-by-side replicates per cell.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |