Lecture 33

IIT Kanpur July 201
20 followers
Mar 1, 2024
hello friend welcome to the course
multivariate procedure with r so you can
recall that in the last three lectures
we have considered the test of
hypothesis related to the mean from a
normal population so first we consider
the one sample test where there were two
conditions that Sigma square is known or
Sigma square is unknown then this one
sample test was extended to two sample
test so what was the difference in the
one sample test we were interested Only
In the comparison of hypothetical value
of the population mean then in the case
of two sample test we wanted to
compare the population means of of two
populations on the basis of two samples
of
data now can we extend
it we have done from 1 to two now can
you make it more than
two in the last lecture I have given you
the examples that where would you like
to use two sample
test now in this lecture we are going to
extend it to more than two population
that is we want to
compare the population
mean from various normal
populations on the basis
of samples of data means from every
population we have got a
sample now the next question is why
should we do it where it is going to be
useful now in order to explain you let
me try to take a very simple example
think of a the situation or think of a
college
where the students
study from class one to say class
10 now do you think if I want to
understand what is the average age of
the students who are studying in class
one class 2 class 3 class 4 class 5
class 6 7 8 9 10 then what will
happen if in case if I try to compare
the ages of students studying in class 1
and Class 10 do you think that are they
going to be the same certainly
not there will be huge
difference so in case if I want to test
H not mu1 is equal to mu2 where mu1 is
the age of the students in the class one
population and Mewtwo is the age of the
students in the class 10 population
there they are not going to be the same
they will be different but on the other
hand if you try
to compare the means of the age of the
students in class 9 and Class 10 or say
class one and class
two do you think that there is going to
be really big difference in the sample
values or in the population mean of
these two
populations maybe
not there is not much difference in the
ages of a student studying in class 1
and Class 2 or say class 9 or class 10
so now my question is this if I want to
compare the ages or the average ages of
the populations which are coming from
class 1 2 3 4 up to
10 so your null hypothesis which was
earlier s not M1 is equal to to mu2 will
now be extended to H mu1 isal to mu2 is
= to mu3 isal to mu4 isal to mu5 equal
to mu6 = to mu7 = to Mu 8 = to mu9 and
equal to mu10 so you are trying to
compare the population means of 10
different
population so this is achieved by the
test which is called as analysis of
variance
or briefly it call is Ana a n o v a
anoa and when we are trying to conduct
such a
test then there is always an assumption
about the population
variance you can recall that in the last
lecture when we considered the two
sample test we had
assumed that the population variance of
population 1 and two they are unknown
but they are
equal well if they are not equal then
you have a different
procedure so now in the case of analysis
of variance we are going to extend the
two sample test to a more than two
sample test so the same assumption will
continue here that we are interested in
the null hypothesis H not mu1 equal to
mu2 up to mu10 but we are assuming that
the variances of these population they
are
unknown but they are
equal so the question comes whenever you
want to conduct such type of test of
hypothesis on a real set of data first
you need to test the hypothesis about
the equality of the
variances or this is also called as we
need to test the homogeneity of the
variances that
whether all the population variances
they are the same or different and that
is the same thing you have to do in a
two sample test case also mean first you
have to test the hypothesis as not Sigma
1 square equal to Sigma 2 square if that
is accepted then you go for the T Test
if that is not accepted then you have to
go for uh some other
procedure so this is called here is a
Bist
test which conducts the test of
hypothesis for the homogeneity of the
variance or the equality of the
variances so in this lecture we are
going to talk about two topics one is
analysis of variance and another is
Bartlet
test well I'm not going to give you here
much mathematical details but my main
objective is to explain you the
procedure and the main steps and after
that I will try to illustrate it in the
art software so in order to understand
the outcome of the art software it is
important for you to understand how they
have been found and what are they
indicating so with this objective let us
begin our lecture and try to understand
analysis of variant and butlet test
right so in this lecture we are going to
cover two topics analysis of variance
and homogeneity of variances and yeah
they are going to be handled only for
the univariate data so first we try to
consider what is analysis of variance so
this is actually the test of hypothesis
for the equality of means when we have
more than two population means right so
suppose we have more than two samples
say samples which are drawn
independently from each of the
population so we call them SK
independent sample and suppose the
sample one is of size N1 sample two is
of size N2 and in general I can say here
that each sample is of size n i which is
drawn from a normal population with mean
mui which is unknown and a common
variant Sigma square that means Sigma
square is unknown but they are equal for
all the population and the samples are
independent and their respective
population variance are unknown but
equal and we want to examine we want to
test if population means mu1 mu2 mu are
different or they are the same so this
is achieved by the analysis of variance
so essentially we want to conduct the
test of hypothesis H not mu1 = to mu2 =
to Mu
K and the alternative hypothesis here is
at least one mui is different from mu J
right where I is not equal to J equal to
from 1 to K so if any pair of the this
mu UJ is not equal then the hypothesis
will be rejected and the next step after
this will be to find out that which of
the perer is responsible for the
rejection of the
hypothesis so now if you try to
understand what is the idea behind this
analysis of variance so first we try to
understand it that why it is called as
analysis of variance mean earlier it was
called as one sample test two sample
test or but here it is called as the
analysis of variant so we understand
that whenever we are dealing with the
statistical Tool uh we are dealing with
the random
variation right so whenever an
experiment is conducted the variable
will certainly enter into the recorded
data now this variability can be caused
by two types of variation one variation
which is due to
assignable causes that we know and this
variation can be explained no issues I
have no problem once I know that uh this
is the reason for the variability and uh
second factor is random causes this is
the part which is which cannot be
explained and this is the Unexplained
variation so now when we are are trying
to develop any good statistical
procedure we always want that uh yes we
variation is going to be there but I
should know the reason and the random
variation should be as small as possible
right so we expect that in any good
statal procedure the explained variation
is as high as possible and The
Unexplained variation which is due to
the random factor is as low as possible
so if I can partition the total
variation into two parts which is due to
the explained fact fa s and another
which is due to the Unexplained
factors then possibly we can get an idea
about the decisions that whether they
are going to provide us the good
decision or bad decision so the next
question is how to partition the total
variance in the data into two such
components this is the job of analysis
of variance tool and this is achieved
through the analysis of variance so
essentially the analysis of variance
partition the variability in two types
of causes due to assignable causes which
is the explained variation and due to
the random causes which is the
Unexplained variation now the question
is how to partition this total variance
in the data into two such components
which I can say broadly as between
sample and within sample so the
variability will be now divided into two
types of variability between sample
variability and within sample
variability and this is the job of the
analysis of variance
right so in the casee of analysis of
variance which is popularly called as
Anova OVA this works by comparison by
comparing the between sample variability
and within sample variability what is
between sample variability the
variability in the sample means say X1
bar X2 bar XK bar because now you have
got here K independent sample for each
of the sample you can find out the
sample mean so you will have here the K
sample
means
K sample mean and if you try to find out
the variability in the values of K
sample mean that is the between sample
variability and then we have within
sample variability that means in each of
the KE sample there are going to be N1
N2 NK number of units so they will also
contribute to the variability and we can
find out the variability within the
sample so for example in the first
sample there are N1 units so N1 units
will contribute to the within sample
variability so I can say that when the
between sample variability that is here
in the one is much larger than the
within sample variability which is here
in the two what does this indicates that
between sample variability is much much
higher than the within sample variable T
this indicates that the population means
are not equal for example if you try to
see the if I try to take the same
example that we have suppose some
student from class one and we try to
record their ages and there are some
student from say class 10 and we try to
record their ages do you think that the
ages within this class they will be
small the the values of Ages within the
class 10 they will also be small but the
value of the ages between class one and
Class 10 this is going to be larger and
this is going to indicate the average
ages of the students in class one and
Class 10 they are not the same so then
that is what I mean to say that
indicates that the population means are
not equal or I can say that H1 is true
or H not is rejected so this approach
depends on comparing the within and
between group variability so that is why
it is termed as analysis of variance
right variance is for variable T now how
are you going to conduct this test that
is what I want to now explain you so we
would like to find out the total sample
size total sample size is the size total
of say N1 N2 NK that is N1 plus N2 plus
up to NK and then for each of the sample
which is based on the sample size n i we
need to find out the sample mean xar I
and Sample variance SI I Square by these
Expressions right sample mean is
obtained by 1 upon ni I summation XI and
Sample variance is obtained by 1 upon n
minus1 summation x i- xar i whole and
then we try to find out the mean of all
the observation N1 plus N2 plus NK
observation which is here n so let xar
be the mean of all the observation from
all the group based on N observation
this n right now then we try to find out
the between sample variability so this
is found as the variance of the K sample
means weighted by sample size and uh
this Express in terms of mean Square
treatment right or say this is because
we are trying to use here the
terminology between variability and
within variability so I'm trying to
indicate here as Say by here mbv the the
mean Square between variable T so it is
just for the sake of understanding right
and in general you will see that this is
also called as mean Square due to
treatments so it is the mean Square due
to between sample variability is defined
by here ssbv divided by the degrees of
freedom right s say BV is computed by
this expression which is the sum of
squares due to between sample variable T
right so this is actually if you try to
see this is xar I minus xar so it is
trying to consider the variability in
the sample means xar 1 xar 2 xar here K
next we try to find out the within
sample variability so this is found as
the weighted mean of the sample
variances which is expressed as the mean
square error or shortly called as MSE so
this MSE is defined as ssse divided by
the degrees of freedom right so this
ssse is actually here like this
summation I goes from 1 to K N IUS 1 s i
squ this is the sum of a square due to
the random errors right and it gives us
some idea about the variability in the
observation due to the randomness in the
values and you can see here that the
degrees of freedom here are here and
minus I will try to show you that how
these things actually
come right so now after this we try to
use some tools of statal inference yes
I'm not giving you here all the details
and which is out of the our view of this
course also but if you wish you can look
into the books and you will get a
complete idea so now we have to find out
the sum distribution of this msbb and
MSE based on that we are we are going to
create a test statistics for conducting
the test of hypothesis for H not mu1
equal to mu2 equal to Mu K right so I'm
just giving you here the brief step so
that you can understand understand how
different values are coming right so
using the properties of normal
distribution we try to compare the MS BV
with msse right and for that we take the
ratio and that ratio will be called as F
statistic right and in order to find out
the F statistics we consider here the
statistical properties that sum of
square due to between variability
divided by Sigma Square this follows a k
Square distribution with K minus1
degrees of freedom under H the sum of
square due to error divided by Sigma
Square follows K Square distribution
with n minus K degrees of freedom under
H not and uh and both this ssbv and ssse
are independently distributed now we try
to take the ratio of this msbb and MSC
that is sum of squares divided by their
respective degrees of freedom and this
Statistics will follow f distribution
with K minus1 and N minus K degrees of
freedom under H not right and this this
degrees of freedom K minus one this is
indicated by df1 and N minus k s here
df2 right so the rule to T the
hypothesis H uh mu1 = to mu2 = to Mu K
that is mu1 H is mu1 = to mu2 = to Mu K
is the same that reject H not at Alpha
level of significance if P Val is less
than Alpha all right so now this msbb is
also known as the sum of square due to
treatment and it is also indicated as a
mstr in various books and this ssbv is
also known as sum of square due to
treatment and many books you will see it
that it is indicated by SST sometime it
is written as here like this also and
now this total sum of squares total sum
of squares which is here say here sum of
all the observations or the to Total
variability in the system that is
divided into two parts sum of square due
to treatment and sum of square due to
error Orly I can also write then total
sum of squares is equal to say say
Square due to say BV plus SS e so
whatever you want to use they are the
same thing and these quantities are very
conveniently displayed in the form of an
anova table right so so you can see here
so you can see here this is a very
simple table which is trying to compile
all the information at one place so here
the First Column here is source of
variation which is here between groups
and within groups and within group is
actually error and then we have here
total then the next column is about
degrees of freedom So This Ss BV upon
Sigma square that has got a k Square
distribution with K minus1 degrees of
freedom so this K minus1 is indicated
here
right SS upon Sigma square that has got
a k Square distribution with n minus K
degrees of freedom so this n minus K is
given here now similarly we can make a
conclusion about the total also but then
we need not to do it because these
degrees of freedoms are additive so if I
try to add here k minus1 + n minus K
this will come out to here n minus one
right so that is why I have not found
here the different types of expression
for the total now in the next column I'm
trying to give here the sum of squares
so sum of squares due to between
variation it is here indicated by ssbv
or say we have also given the name as
here
SST then we have obtained the sum of
square due to error and now if you try
to add here
ssbv plus SSE then you will get here
total sum of square TSS that is the
reason that I have not given given you
here the expression for finding all the
TSS right and this additivity this is
due to
the
Fisher Cochran theorem any
statistics so this partitioning of total
sum of squares due to between variation
and error it has been done orthogonally
that means both these sum of squares are
orthogonal to each other and uh then
they Square distributions their degrees
of freedoms are also additive additive
so all these things they are insured by
the Fisher CM theorem well I'm not going
into the detail of these things now once
you have obtained the see here sum of
squares and the respective degrees of
freedom then you can obtain here this
mean squares due to between variation as
ssbv divided by K minus one so both
these columns are used here msse is
obtained here say SS upon degrees of
freedom n minus K and now this both msbb
and MSC they are obtain here say msbb
upon MSE and this is here the F
statistic right which is called here A F
value so this is here the analysis of
various table and you will see that in
the software outcome also usually they
try to indicate the outcome in the form
of an analysis of variance table right
so let us try to take some examples and
try to understand how to conduct this
analysis of variance in the art software
suppose we have a class of 20 students
and based on that they have been graded
and the grades are ab and C which are
given on the basis of the marks for
example if there is a rule that somebody
somebody if somebody gets a more than
60% Mark then it then then the person is
classified as first Division if between
45 and 60% Mark then the person is
classified as second division and so on
so similar in in the class these 20
students marks have been converted into
GDs a b and c so these marks are stored
in the data Vector here marks and the
corresponding grds are stored in the
data Vector here grds which are obtained
as as Factor right factor is also a
concept in the r software where we try
to convert the numerical data into
categorical variable which is popularly
called as factor in the r software so if
you try to understand the data is like
this that the first student has got
marks 20 and the person has got the
grade C similarly the second person has
got marks 10 and the second person also
has got has got the gr C and so on right
so now the question is that we want to
know if the average marks in each of the
group AB BC are the same or different
because you you can see here that when
we are trying to give the grd then there
is certain range based on which we try
to give the grades for example here you
can see here that both this student who
have got marks 20 and 10 they have been
given the great C whereas if uh somebody
has obtained the marks here 60 then it
has gone got here the Grade B right and
the and another student who has got here
the marks 40 that person has has also
got the grade here B right so this
happens in all the class that the
grading is done on the alphabets based
on certain ranges of the mark so now we
want to understand whether the average
marks in the group a group b Group C or
grade A Grade B gr C they are the same
or not so in our software uh I will try
to show you here two uh ways in which
you can conduct the analysis of variant
so First Command in the our based
package is a o and Analysis of variance
then you have to give here the formula
mean formula you have to specify uh
these two variables for example here
marks and grades because they are
related in a c way I will try to explain
you then whatever be the data here you
have to give here the data frame in
which the variables are specified in the
formula and uh then you have the option
of here position should be written or
not then here QR here QR is for The
Logical flag that should the Q the
composition be written or not then we
have here contrast right contrast as
some linear parametric function where
the sum of the coefficient is equal to
zero and so on but anyway I'm not going
into that much detail and uh you can you
yourself can see but I want to take care
the minimum commands to uh to make you
understand that how to conduct the
analysis of variant so now the first
step what you can do that if you want to
understand whether there is a difference
in the average marks in the groups ABC
the first option is that you can create
here a box plot group box plot and if
you try to recall we have learned these
things in the beginning and I had told
you that as we move further we will be
using this thing so if you try to see
here these are the box plots of uh
groups a b and c this is for here a this
is for here B and this is for here C and
you can see here that the average that
is the median this middle value is in
these boxes is trying to give you the
second quartile which is the median and
you can see here that they are quite
different so if all these box plots are
very very close to each other like this
one then possibly I can say that the
average marks are quite the same right
but anyway we try to conclude the same
thing which we have concluded from the
graphic through this quantitative method
also so now I try to use here the
command a oov and you can see that how I
am trying to write down here the formula
Marx Kaa graes right and then after that
you get here this type of outcome so you
can see here that these are here
grades and these are here residuals
residuals is something like random
errors in simple way what we have
understood yeah different software will
have different meaning or say different
terminologies now we are trying to find
out here some of squares
so I can see here this is the sum of
squares due to
grades or in your terminology this is
ssbv or this is SS treatment or SS you
have written as SST now after this what
is here this value 21 44.95 to this is
the sum of a square due to
residuals so this is your here
SSE right and then in the second row we
have the degrees of freedom which which
is related to the sum of square due to
between variability BV or SS
treatment and this is here 27 degrees of
freedom related to
SSE right so in case if you want to find
out here the TSS total sum of squares
then this can be obtained as a 1636
2.54 8 + 21
44.95 2 and the degrees of freedom of TS
will be 2 + 27 right and now it is
giving you here the in the this row here
residual standard error which is
89130 61 so this is the value of here
square root of say s Square which is or
the the value of standard error that is
value of the estimated value of the
sigma right now there is a another line
here estimated FX may be unbalanced so
well we are not going into that much
detail about the concepts of the this
analysis of variance so who can safely
ignore this line right and you can see
here this is here the screenshot of the
same thing right and now if you want to
make it here more information here for
example if you want to get about the
value of f statistics Etc then you can
use here the command here summary and
the same command here AO marks Del
grades right so this command will give
you this type of outcome here where you
have here the results about the
statistical inference for this different
sum of squares so you can see here this
is here the degrees of freedom which is
here like this right some we s a line
but I can show you it on the this here a
screenshot yeah I think better you can
consider this is screenshot which has a
nice out outcome then in the second
column we have some of his squares so
this value here is the sum of squares
due to
Gres this is something like
ssbv and this 2145 is the
SSE now this now now in the next column
we have here mean Square which is 81 81
is obtained by here 1 163 63 upon here 2
that is the mean SS PV divided by the
corresponding degree of freedom and then
the 79 this is divide obtained as 2145
upon 27 and then finally we have here F
value right this is here 103 and after
that we have here see here P value this
is your here P
value so now it is trying to give you
here different option that okay the
stars are giving significant code but if
you try to see here if you the P value
here is something like 2.32 into 10
power of here minus33 very close to zero
and if you try to take here Alpha is
equal to
0.055% level of significance then you
can very easily say that P value is less
than Alpha and your rule says that
reject H not when P value is less than
Alpha so reject H not means and you are
trying to have here uh mu a isal to Mu B
isal to Mu C that means that means and
of the grades of group a b and c they
are not the same this is rejected and it
was also clear from the box plot also
that they are very different right so
let me try to show you this thing on the
r console also so that you can get here
more
confidence right so let me try to First
enter this data in the AR console and
then now if I try to uh take here this
command here
say so if I try to show you here the box
plot
so you can see here let me try to make
this window smaller you can see here
this is here the box
plot and if I try to make it here this
analysis of variance so you can see here
you are going to get here the same
outcome here like this and if you want
to have the summary of uh these things
so you can see here this is here like
this uh summary
of this this analysis of variance it is
here like this right so you can see here
it is not very difficult the main thing
here is that you need to know how to
interpret it that is more important the
interpretation is more important than
doing this analysis that is a
straightforward what Command right so
now let me come back to our the slides
and now I would try to give you the idea
about another command to conduct this
analysis of variance well you can use
any one of them so another command to
conduct the this one way andov I will
try to give you idea of what is one way
it is one o n e w y one way.
t and uh so this is going to do the same
thing what we have done in the command a
that it will try to test s not mu1 equal
to mu2 equal to Mu K right and in this
case the variances are not necessarily
assumed to be equal right so that is the
main difference between the two and the
command here is one we do test then you
have to give it the formula then data
then subset then na doaction variance do
equal to false or true so this option is
here then formula is as I told you that
we have used the command here marks kild
great like this so This
LHS rhs LHS gives the sample values and
rhs gives the corresponding groups right
and data is the optional Matrix of data
frame in which you have to give the data
subset is another optional Vector which
is trying to specify the subset of
observation to be used and na do action
is a function which indicate that what
should happen when the data has missing
values na a and variance dot sorry VA
dot V do equal that is the variance is
equal or not that command can be used by
specifying V do equal this is a logical
variable which can take the value uh
true or false right so now if you try to
see on the same data set I try to
conduct this uh test so we have a
command here one way. test marks Kilda
gr that is the most simple option so it
will give you here this type of option
so you can see here it is giving you
here the the value of of
Statistics then the degrees of freedom
of the numerator degrees of freedom of
the
denominator and then here P values here
right so and then you can see here this
is here the a screenshot and here I have
intentionally taken that uh variance do
equal is equal to false well if you try
to take it here variance do equal is
equal to true then you will get the same
result what you have got earlier but I
wanted to show you that what really
happens when you try to make this option
that the variance of different
population are not the same that is
based on a different type of approach
but that has some issues one issue which
you can observe here that the degrees of
freedom here in the
denominator I should use here this here
is coming out me
14865 and the degrees of freedom it is
2. it is coming in in the fraction so we
we have certain procedure by which we we
try to approximate the degrees of
freedom when the variances are not the
same but if you try to use here variance
do equal is equal to true then you will
get the same outcome but my objective
was to show you that what really
happened so before I try to explain you
about the but list t let me try to show
you these things on the our console also
right so let me try to copy this command
here and you can see here that it will
give you here like this
right and if you try to use here the
this command where you have an option V
do equal is equal to false then you can
see here this is the default both the
options are coming out to be the same
right but on the other hand if you try
to use here the option here true then
you can see here this won't happen and
you are getting here the same option
which you have obtain
earlier right so that is why I have
shown you this option over here so that
you can yeah you should know how to
conduct these things because in real
life you will be facing different types
of of scalles and and issues now in the
last let me try to very quickly give you
uh information about the the botet test
which is used for conducting the test of
hypothesis for the equality of different
variances for example you have assumed
in the analysis of variance that you
have got Cas samples which are drawn in
from K different normal population which
are having different means mu1 mu2 mu K
but they are having variances like Sigma
1 square Sigma 2 square Sigma K Square
so in the usual analysis of variance you
assume that all Sigma 1 square Sigma 2
square Sigma K Square they have got the
same value But first you need to test so
that is why when you try to conduct the
analysis of variant first you must
conduct that test of hypothesis about
the homogeneity of variance if that is
accepted then you try try to conduct the
analysis of variance with an option for
example like V do equal is equal to True
otherwise use the option V do equal is
equal to false right so a basic
assumption in this analysis of variance
is that all the samples are come from a
k normal population which are
independent and they have got different
means but the same variance so in order
to check whether they have got the same
variance or not we use the Bartlet test
and it test that if Case samples have
equal variances across the sample and
this is called as homogeneity of
variance right so basically we want to
test the null hypothesis H Sigma 1 sare
= to Sigma 2 sare equal to dot dot dot
up to Sigma K square and my alternative
is that at least one Sigma I square is
different from Sigma J Square for J not
equal to I I goes from 1 to K right so
now in this case the test statistics is
here like this well it is only for
information we are going to use the
software and the pool variance that we
used that use in the case of two samples
if you recall this can be computed in
this case by SP Square which is here
like this the same definition has been
extended and the pool variance is
defined as the weighted average of the
group variances so now we have here k
group so each of the group has got the
variance ni minus 1 SI squ so that is
now taken the
average right so in order to conduct
this Bartlet desktop on the data in the
our software we have a command here
Bartlet do t b r t l e t. t
s all in lowercase alphabets and then
the formula here is given by the same as
we did in the case of analysis of
variant marks TAA grades and once you do
it actually this is available in the
base package of R once you try to do it
it will give you here the value of the b
list
statistics here like this 1.1 382 which
has got two degrees of freedom and its P
value here is
0.566 so now based on this P value you
can take appropriate decision and then
you can conduct the test of hypothesis
with the assumption that whether your
variance should be equal or not now I'm
sure that you know about it and let me
try to show you this butlet test in the
our software so you can see here it is
here coming like this
right so this is the same thing which I
shown you on the in the slide right okay
so now we come to an end to this lecture
and you can see here that we have
considered here the analysis of variance
and Bartlet test for the for uh testing
the homogeneity of the
variances now you can see here the
analysis of variance what I have
considered here this is actually the
one-way analysis of variance well
analysis of variance itself is a whole
subject so when we have only one
factor which is affecting the outcome
then we consider one way analysis of
variance when we have two factors then
we try to consider two-way analysis of
variance Etc but definitely we are not
going into the part of analysis of
variance but we wanted to use the tool
of analysis of variance for comparing
the population from say more than two uh
different
populations right and
uh now you have seen that we have
considered earlier one sample test two
sample test for the equality of the for
testing the equality of the population
mean but now we have extended it to more
than two
populations the next question comes here
if your null hypothesis is rejected that
means all mu1 mu2 muqa they are not the
same then the next uh aspect is that how
to find out that which of the this mu1
M2 M they are same and which are
different so this is achieved by the
multiple comparison
test so I'm not going to cover here but
this is for your information only and
most of this multiple comparison test
they are available in the r software so
I cannot cover here each and everything
otherwise the whole this course will be
based only for this analysis of varans
but my very humble request to you all is
that you please try to look into the
books you have a concept of oneway
analysis of variance two-way analysis of
variance different types of multiple
comparison test Etc and you simply have
to just see what they are trying to do
and you have to Simply find out the
command by which you can execute them in
the art software now you can see the
conducting of test of hypothesis is not
difficult in the art software you simply
have to look into the
value of the P that is P value but the
more important part is which of the test
statistics is going to be used whether
it is z test T Test or something else
that is the point which you have to
learn so there are many many parameters
different types of hypothesis uh which
can be tested using different types of
Statistics but surely my scope here is
limited and uh now my objective is how I
can extend it to a multivarious setup so
you please try to revise at least one
sample test two sample test and uh I
have not covered here the confidence
interval estimation because in the uh I
have explained you how to construct the
confidence interval estimation in the
case of one sample in the two sample
also it can be done and in the analysis
of variance also some of the multiple
comparison test like Shi test 2K test
Etc they are based on the confidence
interval so but but now I believe that
you have that you have now sufficient
background so that you can learn
yourself also and I'm sure that it won't
be difficult for you and you will do a
very good job so you try to take
different data sets try to see what is
happening try to expose them to the art
software and try to analyze the outcome
and try to understand how can you
interpret it the better you can
interpret it the better you are so you
try to practice it and I will see you in
the next lecture till then goodbye
by
