A spreadsheet to evaluate SAS Type IV sums of squares

Authors: Klasson, K Thomas

Submitted to: Software and User Manual Public Release
Publication Type: Other
Publication Acceptance Date: 5/28/2021
Publication Date: 5/28/2021
Citation: Klasson, K.T. 2021. A spreadsheet to evaluate SAS Type IV sums of squares. Software and User Manual Public Release. https://doi.org/10.1021/scimeetings.1c00598.
DOI: https://doi.org/10.1021/scimeetings.1c00598

Interpretive Summary:

Technical Abstract: Type IV Sums of Squares (SS) were introduced by SAS in 1976. This method is based on creating testable hypotheses that make sense to the experimenter who intended to use a Type III approach but had experimental failures that resulted in loss of entire cells for some treatment combinations. The method has been criticized but it is still used and available in SAS. The problem is that Type IV SS for SSA and SSB depends on the order of the data entry structure (e.g., which is the first, second, third row, etc.). It also depends on the size of the cells. Specifically, it depends on which is the last row (or column) and the sizes on the cells in that row (or column) in the Cell Means table. In some cases, depending on the data entry structure, some cells that contain data are ignored in the hypothesis testing. Type VII Sums of Squares were introduced by Klasson at the Virtual 2021 ACS Spring Meeting as a method to improve the Type IV SS created by SAS by reordering the rows and columns in the means table so that the hypotheses tested take advantage of all cells. One requirement is that at least one row and one column have all the cells filled. If this requirement is not met, the hypotheses developed and tested are the same as in Type IV. This spreadsheet was part of the presentation at the ACS meeting.