Boolean Factor Analysis

Boolean Factor Analysis (8M)
The data for this example came from a study of lymphocyte blood cells of 50 patients tested with a panel of 20 reagent antisera (vaccines). The underlying assumption is that the cell surface has features that can be recognised by constituents in the reagents. A reagent will cause a reaction if the cell tested has any one of the features associated with the reagent. Cells from a given person are assumed to have a few of these distinct surface features.

In this example we state that a reaction (Response) is coded as 6 or 8 in the dataset and that no reaction (Nonresponse) is coded as a 1 or 2. Any other value in the data is now considered to be an unknown reaction.

Based on some initial analysis, we know that there are 36 positive reactions and 6 negative. It is possible to reduce the subsequent number of discrepancies by:
1. Reducing the number of initial factors
2. Increasing the total number of factors

The graphical output produced by BMDP displays two matrices, (i) the Data matrix and (ii) the Discrepancy matrix. The Data matrix displays the dataset as recoded into response, no response and unknown. The Discrepancy matrix displays positive, negative and unknown discrepancies in matrix format.