Factorial design
Experimental Design
Experimental design refers to the framework or structure of an experiment and as
such there are several experimental designs. We can classify experimental designs
into two broad categories, viz., informal experimental designs and formal
experimental designs. Informal experimental designs are those designs that normally
use a less sophisticated form of analysis based on differences in magnitudes, whereas
formal experimental designs offer relatively more control and use precise statistical
procedures for analysis.
Important experiment designs are as follows:
(a) Informal experimental designs:
(i) Before-and-after without control design.
(ii) After-only with control design.
(iii) Before-and-after with control design.
(b) Formal experimental designs:
(i) Completely randomized design (C.R. Design).
(ii) Randomized block design (R.B. Design).
(iii) Latin square design (L.S. Design).
(iv) Factorial designs.
Factorial design
Factorial designs are used in experiments where the effects of varying more than one factor are
to be determined. They are specially important in several economic and social phenomena where
usually a large number of factors affect a particular problem.
Factorial designs can be of two types:
(i) simple factorial designs
(ii) complex factorial designs.
(i) Simple factorial designs: In case of simple factorial designs, we consider the effects of varying
two factors on the dependent variable, but when an experiment is done with more than two
factors, we use complex factorial designs. Simple factorial design is also termed as a ‘two-factor-
factorial design’, whereas complex factorial design is known as ‘multifactor- factorial design.’
Simple factorial design may either be a 2 × 2 simple factorial design, or it may be, say, 3 × 4 or 5
× 3 or the like type of simple factorial design. We illustrate some simple factorial designs as
under:
Illustration 1: (2 × 2 simple factorial design).
A 2 × 2 simple factorial design can graphically be depicted as follows:
Experimental variables
Control variability Treatment A Treatment B
Level I Cell 1 Cell3
Level II Cell2 Cell4
Complex factorial designs
Experiments with more than two factors at a time involve the use of complex factorial
designs. A design which considers three or more independent variables simultaneously is
called a complex factorial design. In case of three factors with one experimental variable
having two treatments and two control variables, each one of which having two levels,
the design used will be termed 2 × 2 × 2 complex factorial design
which will contain a total of eight cells as shown below in Fig.2
Experimental Variables
Treatment A Treatment B
Control Control Control Control
variable variable variable variable
2 LevelI 2 LevelII 2 LevelI 2 LevelII
Control LevelI Cell1 Cell3 Cell5 Cell7
Variable
Level II Cell2 Cell4 Cell6 Cell8
1
Advantages
They provide equivalent accuracy (as happens in the case of experiments with only
one factor) with less labour and as such are a source of economy. Using factorial
designs, we can determine the main effects of two (in simple factorial design) or more
(in case of complex factorial design) factors (or variables) in one single experiment.
They permit various other comparisons of interest. For example, they give
information about such effects which cannot be obtained by treating one single factor at
a time.
The determination of interaction effects is possible in case of factorial designs.
BLOCKING AND CONFOUNDINGSYSTEMS FOR
FACTORIAL DESIGNS
Blocking is a specialized factorial design in which blocks are included to reduce variability,
in statistical comparisons.
The blocking technique is applied to industrial level experiments, though it is
doubtful whether variability (really) exists or infuences. Blocking improves the precision of
the responses, by reducing/eliminating the variability that is transmitted from:
Extraneous factors or nuisance factors,
Factors, unknown or uncontrollable
Sources, known and controllable.
Normally, materials and processing conditions remain same in all the runs (in a
DoE), then, the result (response) becomes reliable and significant. Else blocks are
constructed to make the experimental conditions homo-geneous, to obtain reliable results.
(a) Change in raw materials supplies: The same API is procured (by the researcher) from a
company manufactured from two batches or two companies or two different processes.
Though there may not be any differences, but their effects need to be evaluated. Such batch
materials (new) are kept in one block (homogenous),because the variability within the block
becomes low compared to those present between batches. Such a situation can be handled
during the planning stage, so that their interpretation can be done appropriately.
(b)Nuisance or bias control : Nuisance factor is an unwanted variable for the researcher, but
affects the variable of interest, e.g. time and day of the study and room-temperature, during
the study. In randomized block designs, blocks are arranged to reduce the bias. Similarly, in
DoE, one factor (of interest) is studied in presence of other nuisance factors. Such factors are
compared against the factor (under study).:
Arranging blocks :Higher order interaction terms, having a '-1' sign, are or grouped as a block
(e.g. the signs of Xg, XXy, XX,X, etc.) (Table1). The DoE is divided into two blocks. The
block l rows are shown in a dark-shaded open circles at corners (on the top of the box) and
block 2is open circle corners (Figure 1).
In 22 design, 4 runs are required. These can be divided into two blocks (2 runs in a
block) (Figure 1a). In the geometric figure, blocks are differentiated by filled and unfilled
circles.
In 23 design, 8 runs are required. These can be divided into two blocks (4 runs in a
block). One block is allotted to the new batch of drugs or excipients. The blocks are indicated
by a filled(dark-shaded) circle and unfilled (open) circles, in the geometricrepresentation
(Figure 1b).
Table1
Two Blocks of Factorial Design -22 DoE-Matrix
Exp.No. Treatment Level of X1 Level of X2 Level of Block
combinatio X1X2
ns i.e.(3)x(4)
(1) (2) (3) (4) (5) (6)
1 1 -1 -1 +1 1
2 X1 1 -1 -1 2
3 X2 -1 1 -1 2
4 X1X2 1 1 +1 1
Wherever X1X2 codes are negative, they are chosen as block1.Rest is chosen as block 2.
Table2
Two blocks of Factorial design-23 DoE-Matrix
Exp.No Treatm Level Level Level X1X2X3i Block
. ent of X1 of X1 of X1 .e.(3)x(
combi 4)x(5)
nation
s
(1) (2) (3) (4) (5) (6) (7)
1 1 -1 -1 -1 -1 1
2 X1 1 -1 -1 +1 2
3 X2 -1 1 -1 +1 2
4 X1X2 1 1 -1 -1 1
5 X3 -1 -1 1 +1 2
6 X1X3 1 -1 1 -1 1
7 X2X3 -1 1 1 -1 1
8 X1X2X3 1 1 1 +1 2
Wherever X X X are codes negative, they are chosen as block 1.Rest is chosen as block2.
The estimation of block effects has made the 2V factorial designs popular and successful.
In the software, provision is made to take care of blocking effects. The software also
suggests which runs are to be grouped as blocks.
When blocks are constructed, the analysis (or interpretation) cannot distinguish
the higher-order interactions(s). Such a phenomenon is known as confounded or aliased.
The X1X2 are confounded (Table 3) and X1X2X3 are confounded (Table 3b). Therefore,
the final effect is a sum of both effects (blocking and higher-order interaction).
Disadvantages : If V (factor) is large, it is not possible to conduct treatments in blocks.
Higher-order interactions are usually more difficult to analyze. It gives a low degree of
accuracy.
CONFOUNDING
Confounding is a modified factorial design in which blocks consist of runs (experiments) to
arrange high-order interactions indistinguishable from main effects,
In full factorial designs, if the number of factors is high, the number of runs increases.
The number of runs can be decreased, by adapting half-fractional/one- -fourth fractional factorial
designs. In that case, it is difficult to accommodate runs (blocks) for combinations. Hence,
incomplete block designs (e.g. balanced incomplete block design, BIBD) are preferred. Thus,
study designs have become the basic layouts on which DoE (this chapter) is placed for
evaluation. When factors are confounded, it is not possible to separate the main effects (X3) from
the interaction effects (X1 X2). However, based on the magnitude of coefficients, the precision of
analysis can be decided.
The confounded coefficients are, C3 and C12. If the C12 is small compared to C3, then,
confounding is less significant. If prior knowledge is available on interactions (and is found to be
less),then, main effects can be identified clearly.
When the codes (signs) of main effects such as l = 23, 2 = 13 and3= 12, then, main
effects are confirmed as confounded.
Higher-order confounding is possible, when the number of factors is large, as in
fractional factorial designs (2 V-2)(Table 4). Columns X3,= X4, have same codes (minus signs, i.e.
same formulation). There fore, it is not possible to estimate factors 3 and 4 separately, because
factors 3 and 4 are said to be confounded (or aliased). Confounding should be used judiciously
applied, only when less important factors are confounded. Hence, the experimenter has to
plan appropriately.
Table4
Codes and confounding 2V-2 design
respons X0 X1 X2 X3 X4
e
+ - - - -
+ + - - -
+ - + - -
+ + + - -
