2. Incomplete Block Designs
๏ฑSome experiments may consist of a large number of treatments
and it may not be feasible to run all the treatments in all the
blocks. Designs where only some of the treatments appear in
every block are known as incomplete block designs.
๏ฑ A balanced incomplete block design (BIBD) is an incomplete
block design in which any two treatments appear together an
equal number of times.
-
3. ๏ฑSuppose that there are ๐ treatments and that each block can hold
exactly ๐(๐ < ๐) treatments. A BIBD may be constructed by taking
blocks and assigning
๐
๐
different combination of treatments to each
block.
๏ฑIn a BIBD setup, each block is selected in a balanced manner so that
any pair of treatments occur together the same number of times as
any other pair.
๏ฑOne way to construct a BIBD is by using
๐
๐
blocks and assigning
different combination of treatments to every block.
4. Examples
๏ฑConsider three treatments, A, B, and C where two treatments are run in every
block. There are
3
2
= 3 ways of choosing 2 out of three. Thus using three blocks
๏ฑ Consider five treatments, A, B, C, D, and E where 3 treatments appear per block.
We use 10 blocks
5. Statistical Analysis of the BIBDโs
๏ฑAssume that there are ๐ treatments and ๐ blocks, and that each block contains ๐
treatments, and each treatment occurs ๐ times in the design (or is replicated
๐ times). Therefore there are ๐ = ๐๐ = ๐๐ total observations.
๏ฑThe number of times each pair of treatments appears in the same block is
๏ฑThe parameter ๐ must be an integer.
๏ฑIf ๐ = ๐, the design is said to be symmetric
6. ๏ฑIn a balance incomplete block design;
-๐ blocks have the same number of ๐ treatment each and
-every treatment is replicated ๐ times in the design.
-Each treatment occurs at most once in block, i.e ๐๐๐ = 1 or 0 where nij is the
number of times the ๐ โ ๐กโ treatment occurs in ๐ โ ๐กโ block, ๐ =
1,2, โฆ . . , ๐ ๐ = 1,2, โฆ โฆ . . , ๐
Incomplete: cannot fit all treatment in each block
Balance: each pair of treatment occur together ๐ times
7. ๏ฑThe Statistical model for the BIBD is
๐ฆ๐๐ = ๐ + ๐๐ + ๐ฝ๐ + ๐๐๐
๐ = 1,2, โฆ โฆ . , ๐
๐ = 1,2, โฆ โฆ โฆ , ๐
Where ๐ฆ๐๐ is the ๐๐กโ
observation in the ๐๐กโ
block,
๐ is the overall mean,
๐๐ is the effect of the ๐๐กโ treatment,
๐ฝ๐ is the effect of the ๐๐กโblock, and
๐๐๐ is the ๐๐ผ๐ท (0, ๐2) random error component.
๏ฑWe partition the total sum of squares in the usual manner; into sum of square due to
treatments, blocks, and error.
๏ฑBut the difference here is that the sum of squares due to treatments needs to be
adjusted for incompleteness.
8. ๏ฑTotal variability may be partitioned into
๏ฑThe total variability in the data is expressed by the total corrected sum of
squares:
9. ๏ฑThe block sum of squares is
Where ๐ฆ.๐ is the total in the ๐๐กโ block. ๐๐๐ต๐๐๐๐๐ has ๐ โ 1 degrees of freedom.
๏ฑThe adjusted treatment sum of squares is
Where is the adjusted total for the ๐๐กโ treatment, which is computed as
with ๐๐๐ = 1 if treatment ๐ appears in block ๐ and ๐๐๐=0 otherwise
10. ๏ฑThe adjusted treatment totals will always sum to zero.
๐๐๐ก๐๐๐๐ก๐๐๐๐ก๐ (๐๐๐๐ข๐ ๐ก๐๐) has ๐ โ 1 degrees of freedom.
๏ฑThe error sum of squares is computed by subtraction as
and has ๐ โ ๐ โ ๐ โ 1 degrees of freedom.
๏ฑThe appropriate statistic for testing the equality of the treatment effects is
12. Example: Suppose that a chemical engineer thinks that the time of reaction for a
chemical process is a function of the type of catalyst employed. Four catalysts are
currently being investigated. The experimental procedure consists of selecting a
batch of raw material, loading the pilot plant, applying each catalyst in a separate
run of the pilot plant, and observing the reaction time. Because variations in the
batches of raw material may affect the performance of the catalysts, the engineer
decides to use batches of raw material as blocks. However, each batch is only large
enough to permit three catalysts to be run. Therefore, a randomized incomplete
block design must be used. The balanced incomplete block design for this
experiment, along with the observations recorded, is shown
13. The analysis of this data is as follows;
๏ฑ From the data, we have ๐ = 4, ๐ = 4, ๐ = 3, ๐ = 3, and ๐ = 12. thus,
๐ =
๐(๐โ1)
๐โ1
=
3(3โ1)
4โ1
=2
๐ = 2
๏ฑ The total sum of squares is
๏ฑThe block sum of squares is =
1
๐ ๐=1
๐
๐ฆ.๐
2
-
๐ฆ..
2
๐
14. ๏ฑTo compute the treatment sum of squares adjusted for blocks, we first
determine the adjusted treatment totals using
= ๐ฆ๐. โ
1
๐ ๐=1
๐
๐๐๐๐ฆ.๐
๏ฑThe adjusted sum of squares for treatments is computed from
15. ๏ฑThe error sum of squares is obtained by subtraction as
๏ฑThe analysis of variance
16. ๏ฑWe conclude that the catalyst employed has a significant effect on the
time of reaction.
๏ฑNote that, in the analysis that we have described, the total sum of
squares has been partitioned into an adjusted sum of squares for
treatments, an unadjusted sum of squares for blocks, and an error
sum of squares:
That is, ๐๐๐ = ๐๐๐๐(๐๐๐๐ข๐ ๐ก๐๐)+๐๐๐ต(๐ข๐๐๐๐๐ข๐ ๐ก๐๐)+๐๐๐ธ
๏ฑSometimes we would like to assess the block effects. To do this, we
require an alternate partitioning of SST, that is,
๐๐๐ = ๐๐๐๐ + ๐๐๐ต(๐๐๐๐ข๐ ๐ก๐๐) + ๐๐๐ธ
Here ๐๐๐๐ is unadjusted.
17. ๏ฑIf the design is symmetric, that is, if ๐ = ๐, a simple formula may be obtained for
SSBlocks(adjusted). The adjusted block totals are
=๐ฆ.๐ โ
1
๐ ๐=1
๐
๐๐๐๐ฆ๐. ๐ = 1,2, โฆ โฆ โฆ . ๐
๏ฑThus, the ๐๐๐ต ๐๐๐๐ข๐ ๐ก๐๐ is given as
20. ๏ฑA summary of the analysis of variance for the symmetric BIBD inclufing both
treatments and blocks
๏ฑNotice that the sums of squares associated with the mean squares do not add to
the total sum of squares, that is,
๏ฑThis is a consequence of the nonorthogonality of treatments and blocks.
21. ๏ฑThe estimation of the BIBD,s model parameters
๏ฑConsider the previous BIBD example.
๏ฑRecall: