Modification of Malinowski's F‐test for abstract factor analysis applied to the Quail Roost II data sets

Malinowski's F‐test is a popular method for selecting the number of factors in abstract factor analysis. Very recently we introduced a modification of this F‐test that is based on an alternative number of degrees of freedom. In this paper the modification is applied to the Quail Roost II data sets. These artificial data sets have recently been included in a data base of reference data sets for chemometrical method testing. Balanced scaling is applied as a preprocessing step, since the noise in the raw data violates the requirement of a constant variance. In comparison with the original test, highly improved pseudorank estimates are obtained by using the modification. These results demonstrate the potential usefulness of the modified F‐test for the analysis of real data, since the noise structure of the Quail Roost II data sets is, for example, very similar to the hypothesized noise structure of data collected from hyphenated instruments. © 1997 John Wiley & Sons, Ltd.