One month ago, you helped me get straight with the four-way ANOVA with mixed balanced design AxBxC(D). Here A, B, and D are fixed, while C (usually subject) is random.
Since now I don't have access to MINITAB, I would have to try a Matlab function 'anovan' first. As far as I understand it, anovan currently only allows plain fixed effects. I am not so sure if this is a concern for a mixed design such as AxBxC(D). I wonder, as long as I get the sum of squares based on the following formulas as though I am dealing with a crossed design of AXBXCXD, it would be fine, correct?
D C(D) = C + CD A AD AC(D) = AC + ACD B BD BC(D) = BC + BCD BA BAD BAC(D) = BAC + BACD Error [formally, R(ABC(D)).]
Basically my concern is, does the fact that C is a random factor in the mixed design AxBxC(D) have any effect on the above calculation based on the imaginary crossed design AXBXCXD. I think that it doesn't, but I need some confirmation before I continue.
Another issue here is, if random factor C has k levels in the mixed design AXBXC(D), I sould still treat C with k levels in the imaginary crossed design AXBXCXD as though C were not nested within D, right?
After I get all the sum of squares, I think that I can construct the F statistics for effect tests myself by using the rules of thumb for writing the ANOVA table without really relying on anything (such as F values) from the Matlab function 'anova'. Instead, 'anovan' would be just a tool to help me get those SSs. Am I on the right track?
Thanks from, Gang
Donald Burrill wrote:
First, a correction about two-way ANOVA with factor B nested within factor A, that is B(A) in the design notation with which I am familiar. It is not correct to describe it as reducible "to a series of one-way ANOVAs", because the one-way ANOVAs will not (in general) have removed the other factor from the error SS. It may reasonably be said to be reducible FROM a two-way ANOVA with crossed factors, since (as you correctly indicate) the SS for B(A) is equal to the sum of the SS for B and the SS for AB in the crossed design. (The corresponding degrees of freedom behave similarly.) Thus if you run a 2-way ANOVA you will get the following sources of variation: A B AB Error (within cells) (formally equivalent to R(AB) where R is the "replication" factor nested within the smalles cells of the design: what one writer has called "the ubuquitous nested factor", but which software often labels merely "Error" and you merely need to add together the SS (and DF) for B and AB to yield the SS and DF for the nested design: A B(A) Error
For a four-way design AxBxC(D) the corresponding result is that the nested effect C(D) is equivalent to (C + CD) everywhere in the design; so you run a standard 4-way crossed design, just as though AxBxCxD, and produce the summary table thus: D C(D) = C + CD A AD AC(D) = AC + ACD B BD BC(D) = BC + BCD BA BAD BAC(D) = BAC + BACD Error [formally, R(ABC(D)).]
If (as is often the case) the nested factor C is a random factor, the rest being fixed factors, the mean squares for C(D), AC(D), BC(D), and BAC(D) serve as the denominator mean squares (aka "error mean squares") for the immediately preceding terms in the table as displayed above. What MINITAB labels "Error" is the error term for C(D) and its interactions (presuming that A, B, and D are fixed factors; if one or more of them are random, things get a bit more complicated).
You may find it convenient to consult any good text on design (e.g., Keppel, "Design and Analysis"; or Glass & Stanley, "Statistical Methods in Education and Psychology" (Chapter 18, particularly sections 18.4 et seq., "Rules of thumb for writing the ANOVA table")) on how to set up the ANOVA table (and carry out computations) for any complete balanced design.
Nested designs may be analyzed fairly easily in MINITAB, if you have access to it: one merely specifies the design in a conventional kind of way (MINITAB conventions, of course, but they're fairly clear). For the design above, a suitable MINITAB command would be: ANOVA Y = A ! B ! C(D) ! D ; random C.
In the ANOVA design statement, "!" means "construct all possible interactions involving these factors"; and MINITAB will refrain from trying to construct impossible interactions like DC(D) . If you want to verify the logic you think you need, ask MINITAB to produce a table of the expected mean squares by including the subcommand "ems" (for "expected mean square"), thus: ANOVA Y = A ! B ! C(D) ! D ; random C; ems.
HTH. -- DFB.
On Wed, 22 Oct 2003, Gang Chen wrote:
I need to run a four-way ANOVA with one factor (i.e., subject) nested within another factor. I did a little research on the funtion 'anovan' in Matlab, and it seems that 'anovan' would not allow me to do nesting directly. My understanding about two-way balanced ANOVA with nested design is that it could be reduced to a series of one-way ANOVA's since
nested SSB(A) = crossed (SSB + SSAB)
Does such a reduction still hold for four-way ANOVA with nested
design? I mean, can I reduce a four-way balanced ANOVA to a few
three-way ANOVA's? If true, how? And, does anybody happen to have
such a Matlab program available for N-Way ANONA with nested design?
----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
-- Gang Chen SSCC/DIRP/NIMH National Institutes of Health Building 10, Room 1D80 10 Center Drive, MSC1148 Bethesda, MD 20892-1148 Phone: 301-594-9193; Fax: 301-402-1370 http://afni.nimh.nih.gov/ssc/
. . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
