Unit-7

Comparison Between Treatment Means

 There are many ways to compare the means of treatments tested in an experiment.
 Only those comparisons helpful in answering the experimental objectives should be
thoroughly examined. e.g if your interest is yield , you have to focus on it rather
than other variables.
 To be understood this, we will see the following example in detail

 Consider an experiment in rice weed control with 15 treatments-4 with hand

weeding, 10 with herbicides, and 1 with no weeding (control).

 The probable questions that may be raised, and the specific mean comparisons that
can provide their answers, may be:
 Is any treatment effective in controlling weeds? This could be answered simply by
comparing the mean of the non-weeded treatment (CONTROL) with the mean of
each of the 14 weed-control treatments.

 Are there differences between the 14 weed-control treatments? If so, which is

effective and which is not? Among the effective treatments, are there differences
in levels of effectivity? If so, which is the best?
 Pair comparisons

 Pair comparison is the simplest and most commonly used comparison in

agricultural research.

The two most commonly used test procedures for pair comparisons in
agricultural research are the least significant difference (LSD) test which is
suited for a planned pair comparison, and Duncan's multiple range test
(DMRT) which is applicable to an unplanned pair comparison.
 Least Significant Difference Test (LSD)
 The least significant difference (LSD) test is the simplest and the most commonly
used procedure for making pair comparisons.
 The procedure pro­vides for a single LSD value, at a prescribed level of significance,
which serves as the boundary between significant and non-significant differences
between any pair of treatment means.
 That is, two treatments are declared significantly different at a prescribed level of
significance if their difference exceeds the computed LSD value.
 The LSD procedure is only used when the treatment source of variation is found to
be significant by the F-statistic (ANOVA table).
 The purpose of testing each mean difference is to test the null hypothesis that the
corresponding two treatment means are equal.
 If the LSD test must be used, apply it only when the number of treatments is not too
large-less than six. OTHER YOU have to use Duncan’s!
 The procedure for applying the LSD test to compare
any two treatments
 STEP 1. Compute the mean difference between the ith and the jth treatment as:

where Xi, and Xj are the means of the ith and the jth treatments.

 STEP 2. Compute the LSD value at a level of significance as:

where Sd is the standard error of the mean difference and ta is the tabular t value,
from Appendix C, at a level of significance and with n = error degree of freedom.

where r is the number of replications that is common to both treatments in the pair
and S2 is the error mean square in the analysis of variance.
 Then LSD can be computed using:

 Example: A data from a CRD experiment with seven treatments (six insecticide
treatments and one control treatment) were tested in four replications. Assume that
the primary objective of the experiment is to identify one or more of the six
insecticide treatments that is better than the control treatment. To do so, we need
the following procedures:
Example: 1. Grain Yield of Rice Resulting from Use of Different Insecticides for the Control
of insects, from a CRD Experiment with 4 (r) Replications and 7 (t) Treatments

Analysis of Variance (CRD with Equal Replication) of Rice Yield Data

 Step 1. ANOVA must be significant, Compute the mean difference between
the control treatment and each of the six insecticide treatments, as shown.

Comparison between Mean Yields of a Control and Each of the Six Insecticide Treatments,
Using the LSD Test
 Compute the LSD value at a level of significance as:

 For our example, the error mean square is 94,773 (from table above), the error
degree of freedom is 21, and the number of replications is 4. The tabular t values
(Appendix C), with n = 21 d.f, are 2.080 at the 5% level of significance and 2.831 at
the 1%. Therefore, LSD is
 Step 3. Compare each of the mean differences computed in step 1 to the LSD
values computed in step 2 and indicate its significance with the appropriate asterisk.


Would it be possible to compare, let say treatment 5 and with other treatments
and so on?
 Duncan's Multiple Range Test
 For experiments that require the evaluation of all possible pairs of treatment means,
the LSD test is usually not suitable.

 This is especially true when the total number of treatments is large. In such cases,
Duncan's multiple range test (DMRT) is useful.

 The procedure for applying the DMRT is similar to that for the LSD test, DMRT involves
the computation of numerical boundaries that allow for the classification of the
difference between any two treatment means as significant or non-significant.

 However, unlike the LSD test in which only a single value is required for any pair
comparison at a prescribed level of significance, the DMRT requires computation of a
series of values, each corresponding to a specific set for pair comparisons.
 To look on DMRT comparisons, we will see the detail using this example
 Step 1. Rank all the treatment means in decreasing (or increasing) order. For yield
data, means are usually ranked from the highest-yielding treatment to the lowest-
yielding treatment.
 For our example, the seven treatment means arranged in decreasing order:
 Compute the (t- 1) values of the shortest significant ranges as:

where t is the total number of treatments,

Sd= standard error of (error mean square) the mean difference
rp =values are the tabular values of the significant studentized ranges obtained from
Appendix F,
p= is the distance in rank between the pairs of treatment means to be compared (i.e., p = 2
for the two means with consecutive rankings and p = t for the highest
and lowest means).
 For our example, the rp values with error d.f. of 21 and at the 5% level of
significance are obtained from Appendix F as:
 The (t - 1) = 6 Rp. values are then computed:
 Step 4. Identify and group together all treatment means that do not differ
significantly from each other:
 For our example, the difference between the largest R, value (the Rp value at p = 7) of
513 and the largest treatment mean (T2 mean) of 2,678 is 2,678 - 513 = 2,165 kg/ha.
 From the array of means obtained in step 1, all treatment means, except that of T3,
are less than the computed difference of 2,165 kg/ha. Hence, they are declared
significantly different from T2.
 Then, the difference between the second largest Rp value (the RP value at p = 6) and
the second largest treatment mean (T3 mean) is computed as 2,552 - 508 = 2,044
kg/ha.
 Because the means of treatments T5, T6, and T7 are less than 2,044 kg/ha, they are
declared significantly different from the mean of T3 .
 The process is continued with the third largest treatment mean. The difference b/n
the third largest treatment mean (T4 mean) and the third largest Rp value (RP value at
p = 5) is computed as 2,128 - 499 = 1,629 kg/ha.
 Only the mean of treatment T7 is less than the computed difference of 1,629 kg/ha.
Thus, T7 is declared significantly different from T4.
 The same process can be continued with the fourth largest treatment mean. But,
because the mean of T7 is the only one outside the groupings already made, it is
simpler just to compare the T7 mean, using the appropriate Rp values, with the rest
of the means (namely: T1, T5, and T6). These comparisons are made as follows:

As the difference b/n T6 and T7 less than p tabulated, thus, T6 and T7 are declared
not significantly different from each other.
 This result can be presented in two ways:
DMRT for Comparing All Possible Pairs of Treatment Means, from a CRD
Experiment Involving Seven Treatments.