We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Understanding Accounting Changes in an Efficient Market: Analysis of Variance Issues.
- Authors
McKeown, James C.
- Abstract
An examination of analysis of variance, Anova design, which includes a blocking factor, must start with an understanding of the effect of the blocking factor on the dependent variable the cumulative abnormal return or CAR. The design includes a separate mean for each block. The effective value of each observation's CAR is thus transformed into the deviation of the original CAR from the mean CAR of the two firms control and switch in the pair of firms. The ANOVA then estimates the main effect of switch vs. control, factor A as the difference between the CAR deviations from the pair means. The other main effects and their interact modify these estimations to allow for the possible effect of signs of the forecasted change in earnings. A test of this main effect then constitutes a test of the market effect of the decision to switch to lifo. An ANOVA, of course, has only one ESS. ESS is defined to be the error term in an ANOVA. An examination of a standard statistics book would have revealed this statement to be inaccurate. Any ANOVA design that has a blocking factor combined with one more crossed factors will necessarily have more than one error term. Error term is actually ambiguous in an ANOVA. Specifically one error term excluding the between-block sum of squares will be used for analysis of any effect that includes the blocking factor while a second error term including the between-block sum of squares will be needed for analyzing effects that do not include the blocking factor. When a blocking factor is used, the observations from a given block must be placed in cells whose factor levels differ only for the blocking factor.
- Subjects
ANALYSIS of variance; STATISTICS; ACCOUNTING; EXPERIMENTAL design; STATISTICAL correlation; MATHEMATICAL statistics; THEORY
- Publication
Accounting Review, 1987, Vol 62, Issue 3, p597
- ISSN
0001-4826
- Publication type
Article