noc19-mg24-Lecture 28 - ANOVA - part 3

IIT Kanpur July 201
20 followers
Dec 3, 2023
[Music]
[Music]
good afternoon students welcome to the
course of advanced green manufacturing
systems we were just going through how
the questions can be used and we are now
going to see how these things are
getting done in practical but remember
that we are interested in going to use
these two equations for calculating the
total values and fill this table so this
was the data that we looked into looked
into the system this is your factor as
we mentioned earlier okay these are our
levels air levels these are our
replications ya n right so in this case
we have what we have what is called as a
is equal to 5 little n is equal to 5
right so you can calculate big n ask air
times little n which is 5 times 5 25
so the begin here is 25 in this regard
okay
I told our ladies we are supposed to
calculate something called why I dot
okay why i dot is the so why I dot is
calculated as I dot is equal to Sigma I
J is equal to 1 to N why I J so for each
value of I for each value 1 so in this
case it is this is given by 7 plus 7
plus 15 plus 11 plus 9 okay that will
give you a value of 49 right then this K
the second value you can do is 12 plus
17 plus 12 plus 18 plus 18 right so that
will give you that is 12 plus this is 40
77 right so you get the value of 77 in
this regard then the this why I dot is
14 plus 18 plus 18 plus
nineteen plus nineteen okay that should
give you eighty-eight okay
and then you have the third one fourth
one was one nineteen plus 25 plus 22
plus nineteen plus twenty-three should
give you one zero eight okay the last
one is seven plus 10 plus 11 plus 15
plus eleven should give you fifty four
right so these are the individual sums
as I said earlier again these are the
some of the rows okay so remember this
is the sum of raw values each individual
row values are summed okay
then I can do something called us y dot
dot okay y dot dot and the y dot dot is
equal to equal to Y 1 dot plus y 2 dot
plus all the way to Y v dot in this case
so I sum this up which is 49 plus 77
plus 88 plus 108 plus 54 that should
give me 376 okay so this is the total
sum okay this is equivalent to summing
all these values okay some should give
you this one right now what the next one
we are going to do here is why I bar
I wrote it as bar so why I dot bar okay
so that is the average of this so this
is calculator does 49 divided by five so
there is one two three four five five
observations right so the why why I dot
bar is equal to Y I dot divided by n
okay so that is the way you do you
divide by the N n is five here right
this will give you the value of nine
point eight okay this is seventy seven
by five okay which is four
fifteen point four in this regard right
the third number is what we calculate us
seventeen point six which is 88 divided
by five then 108 divided by five is
twenty one point six and 54 divided by
five is ten point eight okay so these
are your averages right and then I can
do something called us Y dot dot bar
okay so I can find the in their averages
of it which is y dot dot bar is equal to
Y dot dot divided by n you can do that
so this is in this case it is three
seventy six divided by 25 okay I can get
it the value it comes close to fifteen
point zero four right so this is an
estimate of MU okay
an average it's kind of an estimate of
an average so going through once again
Kuechly first we do the totals these
totals are the row totals okay and what
you do is you sum each individual values
of the row sum them together that will
give you the row totals you do the five
row totals then you do the average this
is the row averages okay and what you do
is you take the sum total divided by the
number of observations in the roughly
how many applications are there you get
this value this is your total grand
total okay this is given by summing
above all the values so I told you how
do we calculate y dot dot you can do
this one or you can calculate all the
values you get it to us three seventy
six and this is y dot bar then you
calculate this how do we do that here is
equation that I told you okay so this is
the first step of analysis okay analysis
step one right then the thing is we use
ANOVA analysis of variance to test
to test h0 okay h0 asked mu 1 equal to
MU 2 equal to mu 3 equal to mu 4 equal
to mu 5 okay we are initially testing
the that so this is your what I call as
null hypothesis okay you belief
believing that all the means the
treatment levels of nog belief on the no
impact on the soldering defects okay and
the alternative hypothesis the
alternative hypothesis okay is h1 what
are we going to do the alternate
hypothesis some means are different so
we are saying that not all of them are
the same some of them have a different
impact on that so if you reject the null
hypothesis okay yeah so Eve rejected
null hypothesis will allow us to believe
this okay
so it's like the same way as please take
a textbook basic probability statistics
book to study hypothesis testing okay
but in this case anyway I am assuming
that you guys have already gone through
it but they always the null hypothesis
the belief is that that the treatment
levels have no impact on the means and
if that belief is not right then we
believe that some of the means are
different that is the alternate that we
are going to take and we already seen
how these numbers are actually
calculated right no second step of the
analysis is okay analysis step two we do
the computations but remember most of
this kind of time you have large number
of data in theory in when you are doing
in a factory you have many number of a
machines and Lord of experiments a lot
of
data that you were collected so doing
this manually is not really an easy
thing but this is a small example
problem so I can do this kind of things
manually to show you how the problem
computation works but in real life you
would use a software program like
sometimes SPSS or jump in or something
like that or things like say Excel to do
this but here we are actually doing this
by hand just to show you the mechanics
of the system alright so first thing we
need to do is we need to calculate the
sum of squares okay yes yes totally
so let's first calculate sum of squares
of total remember in the previous
equation that I told you we calculate
the SS total here and here is an example
of this as total then you covered it SS
treatment from there you calculate
everything else right so going back to
this let's calculate the SS total okay
this SS total is given by the equation
Sigma I is equal to 1 to a sigma j is
equal to 1 to n little n okay so in this
case the a is this is 5 this is also 5
okay and then 5 5 each okay why I J
square okay minus y dot dot square
divided by n this is the equation that
we actually oh we don't need up around
this is I'm sorry okay so this you
square the sums and then you from this
you subtracted which if you elaborate it
it is written as if you go back these
are the values so 7 square plus 7
squared plus 15 square plus 11 square
plus 9 square plus 12 square plus all
the way up to this last 11 square okay
so we do that here 7 square plus 7
square plus fifteen square plus all the
way to 15 square plus 11 square okay so
the first term this part of the term is
this okay it is this one this is the
okay the Ahram tom okay
how did we calculate this is 7 plus 7
plus 15 plus all these values we square
them square them square heads and square
them and this is the last 15 and 11 and
then you sum them up right - okay this
is the first part of the puzzle it's the
first part first time - y dot dot y dot
dot do you calculate it does 376 okay
it's right here 376 okay so we say it
does 376 square divided by n n was
calculated again previously here us
little n was calculated as 24 not little
n big n was calculated as 25 so you
divided by 25 okay
so again the first term is U square each
individual values each individual values
are squared and summed it together right
and that gives you the first term and
then the second path okay he's the
threes
total sums squared divided by 25 and
when you do this number I have
calculated this already so I'll give you
the number six thirty six point nine six
right this is the value you will get
right so please remember this how this
calculation was done right then the next
one we do is sum of squares SS treatment
okay so most course of treatment is
calculated by the equation so we already
told you what the equation is the
previous case 1 over N divided by
multiplied by this equation right so we
are going to do that using here 1 over n
multiplied by Sigma I equal to 1 to a ok
I is equal to 1 to a why I dot square
minus y dot dot square over N right so
this is the equation that we did here
which we are exactly going to do here
the value of a is 5 so we will be doing
this five times
which is equal to one over five little n
is five okay one over five x okay the
white I dot square is the Y i dots are
exactly here so it's 49 square plus 77
square plus 88 square plus 108 square
plus 54 square right so what we are
going to do here is it is 49 square plus
the next one is 77 square plus all the
way to 54 square okay this is the first
term where you are having each one of
these sums you square them and that and
square them and sum them up so this is
what you have as your first term right
okay then - your Y dot dot y dot dot as
I said earlier is 376 right here sum
total office 3376 all square divided by
begin is 25 okay so you have the both
the terms and I get this us for seventy
five point seven six right so this is
the second number I have already
calculated so we have calculated the sum
of squares of total and sum of course
squares of treatments we are done both
right then how do you calculate sum
squares of errors okay
he is calculated by SS total minus SS
treatments so where did I get this I got
this in the matrix remember this the
error sum of squares were calculated us
sum of squares of total divide - M Samus
course of treatment so using that what
I'm going to do it is it is equal to six
thirty six point nine six this is what I
calculated right there - four seventy
five point seven six this is the sum of
squares of treatments okay both of them
put together I get it as one sixty one
point two zero okay so I calculated my
sum of squares of total sum of squares
of treatment and I calculate the sum of
squares of error so what I've done is
I've called
where'd all these AMA squares from the
four so that I can start filling the
table in this regard right okay so these
calculations as I told you earlier
please and remember that these
calculations are always done with the
help of the converter software now we
get into the last step of analysis the
third step of analysis what the first
thing we need to do is we prepare the
ANOVA table okay so you prepare the
ANOVA table where did we see the ANOVA
table this was your ANOVA table right
this is what we call us our ANOVA table
so let us go back and try to prepare the
ANOVA table so I'll just draw this okay
so first thing is we have source of
variation yeah and we have us the source
of variation it was remember this was
between treatments so the main source of
variation for us is the conveyor speed
this is your treatment that is your
source variation then the next one is
you Hardy's error and so if you think
about it it is treatments error and
total right so we go with the same
fashion total right this is our setup
and the second one is sum of squares
okay so where did we get the sum of
squares we have already calculated it
six thirty six point nine six four
seventy five point seven six one sixty
one point two zero so we go back and we
fill those values okay so the first one
is the totally six thirty six point nine
six so six thirty six point nine six I
always write it in this fashion so that
I don't make any mistake then the
treatment is for seventy five point
seven six so it is for seventy five
point seven six
this is your treatment by the way okay
please remember that okay and then the
error is calculated as the difference of
them one sixty one point two zero so it
is one sixty one point two zero right we
draw right this almost question this
regard then we calculate what we call us
the degrees of freedom okay okay so the
degrees of freedom as I said earlier in
the ANOVA table okay degrees of freedom
is a minus 1 n minus a n minus 1 right
so let's do a minus 1 so the value of
the degrees of freedom here is it is a
minus 1 is 5 minus 1 equals 4 so this
value will be 4 okay this is n minus 1
right so if you remember this this is
written as n minus 1 as the degrees of
freedom so n minus 1 is equal to 25
minus 1 is 24 right 24 right and the
this degrees of freedom error degrees of
freedom was written as n minus a right
so if you think about it does n minus a
that is 25 minus 5 equals 20 okay this
value is 20 one other way to check
whether you got the degrees of freedom
correctly is okay the if you look at
these okay these two difference should
be this difference okay remember that
that's another way to think about it you
will calculate this if you take the
difference of the total and that the
treatment degrees of freedom then the
error should also come to that and you
can independently calculate by this
particular equation as well right then
the next one we calculate is what we
call as the mean squares okay so the
mean square is calculated in this case
is mean square treatment okay is given
by sum of squares of treatment divided
by degrees of freedom which is equal to
for seventy five point seven six divided
by four okay and when you do that you
actually get a value one one eight point
nine four okay so the value goes here
one one eight point nine four right then
the next mean square is the mean square
of error okay he's given by some more
squares of error divided by degrees of
freedom which is one sixty one point two
zero divided by twenty which would give
you eight point zero six okay eight
point zero six this value goes right
here right so this is your mean squares
I told you again the mean square is the
sum of squares divided by the degrees of
freedom so ma square is divided by the
degrees of freedom then what we do is we
calculate the F value okay value is
calculated as I told earlier also is
that the F ratio is the means course of
treatment divided by the mean squares of
error right so if you use that the F 0
is equal to M s treat when divided by M
s error which is one one eight point
nine four divided by eight point zero
six you get the value of fourteen point
seven six so this value F zero fourteen
point seven six goes right here right
now that being said now we have
calculated all the major values if you
think about this table we have filled
almost every values right here now the
thing is you have to make a decision out
of this right so the calculated value so
the F value is used to make decisions
okay use the F table in
textbook
and for 95% confidence or alpha equal to
0.05 whatever it is you look at the F
value so you're looking at F of the F is
given by usually calculated as you have
the F of a and error degrees of freedom
okay so that is f of four-and-twenty
these two are our degrees of freedom
goes here for 95% confidence interval is
two point eight seven right this is the
value if you look into the table I would
recommend you to look into it okay so
this degrees of freedom goes by the
degrees of freedom of treatment and the
error okay so this is the treatment this
is the error okay
you calculate both and then you get the
values two point eight seven okay so
what happens here is if you think about
it it is like a probability distribution
something like this and you're saying
the critical value somewhere is two
point eight seven
so anything about this is rejection
reject h0
you are rejecting the null hypothesis so
since fourteen point seven six is larger
okay much larger than two point eight
seven okay we reject h0 we reject the
null hypothesis and conclude that and
conclude that that treatment means
different means different okay so you
are saying that yes as of now because
this value is so large we are saying
that the means are different means are
different in a sense that these conveyor
speeds whatever the speed you select
okay these speeds have a important
significant so the conclusion is that
the speed speed of conveyor
is important okay so you have project Li
you are now identified by one way to
figure out a factor that is significant
means is he say significant factor so
for this soldering machine the conveyor
speed is an important factor that is the
conclusion that you are able to now take
out of this same way you do this for
multiple factors and some of the factors
will come out to become to be
significant some more than will come out
not to be significant so whichever comes
out to be significant those are the
factors that you need to take it outside
for optimization so I hope now you
understand how we selected the factors
that had to be taken into an
optimization model right
so we do we select all the set of
initial set of parameters and then from
there we do what we call as a analysis I
especially collect data do experiments
and then we does the hypothesis that
different treatment levels of this
factor has importance or not importance
and when you can do an ANOVA it will
tell you yes this factor is significant
for you the answer is yes then you take
it forward okay so with this what we do
is we have come to a conclusion of this
analysis of median single factor
analysis of variance please work out the
example and also do the homework problem
because it is important that if you
don't do this practice it then you will
forget this and you will not be able to
do a proper job of understanding this so
please do the practice problems please
do the homework please understand this
is an important technique for you to
learn and please do the needful in the
next class what we will try to do is we
will try to do the factorial experiments
and so that just you have an idea of
which level is more important okay we
are now saying that this factor is
important now the next step will also
tells you which of the levels are more
important for you and that once we
finish that then we will go back and
look into the optimization algorithms
and stuff like that thank you very much
for your patient hearing and I wish you
all the best thank you
[Music]
[Music]
[Music]
