On the design of experiments under spatial correlation (Q3746722)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3982335
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | On the design of experiments under spatial correlation |
scientific article; zbMATH DE number 3982335 |
Statements
On the design of experiments under spatial correlation (English)
0 references
1986
0 references
A-efficient designs
0 references
lattice process
0 references
randomization
0 references
D-pessimal Latin squares
0 references
torus designs
0 references
spatially correlated errors
0 references
ordinary least squares
0 references
A-criterion
0 references
Generalized least squares method
0 references
known correlation matrix
0 references
Knut Vik designs
0 references
knight's move square
0 references
gerechte designs
0 references
0.843658447265625
0 references
0.8350811004638672
0 references
0.8071372509002686
0 references
0.8067902326583862
0 references
0.7946059107780457
0 references
The design aspects of spatially correlated errors are investigated. The theory of analysis of experiments developed is based on the assumptions that there is only one variable of interest, the mn plots are arranged in a rectangular array, and the correlation between adjoining plots is non- negligible in both directions. The methods of estimation discussed are (a) ordinary least squares and (b) generalized least squares. Comparison is made with respect to the A-criterion or \(\Phi_ 1\) criterion which is equivalent to minimizing (i) the average variance of estimated treatment effects (ii) the average variance of an estimated difference in treatment effects and (iii) the expected value of the treatment sum of squares when there are no treatment differences. Generalized least squares method with known correlation matrix is adequately dealt with. The paper considers various Latin square designs, Knut Vik designs and knight's move square of different orders. Optimality is worked out. Examples of pessimal squares and gerechte designs are given. Some special cases leading to more general results are discussed. One is to map a rectangular lattice onto a torus. A number of examples is given. The author has cited a good deal of the literature.
0 references