Facts and Artifacts in Using Analysis of Covariance to “Undo” the Law of Initial Values

ABSTRACT The writer has proposed (Benjamin, 1963) that the criterion for a score linearly independent of initial level be that the score have no correlation with initial level. The criterion makes analysis of covariance (anacova) the method of choice for undoing LIV. This paper reviews miscellaneous artifacts said to be associated with anacova and finds that none of them precludes this procedure. Some presumed artifacts considered are: that rx d is “… so complexly constituted that it does not allow simple interpretation” (Lacey and Lacey, 1962); that anacova can cause a loss of valuable information (Heath and Oken, 1965); that anacova introduces an artifactual association with final level (Heath and Oken, 1965); that LIV as measured by anacova techniques can be an artifact of whether resistance or conductance happens to be chosen (Hord, Johnson, and Lubin, 1964); that the use of anacova in (clinically important) instances where groups are defined by a fixed variable is sure to vitally violate its assumptions (Lubin, 1965).