Evaluation of Textile Materials

Prof Apurba Das

Department of Textile Technology
Indian Institute of Technology- Delhi

Module No.#01
Lecture No. # 03
Sampling Methods and Sample Size

Hello everyone. So we will start the next segment of the course Evaluation of the Textile
Materials which is that first segment it is a Sampling method and Sample size. In this session,
we must understand that what is sampling? What is sample? And why do we want sampling?
(Refer Slide Time: 00:51)

So evaluating or estimating attributes or characteristics of the entire system process or

product through a representative sample the term which is important is representative sample
can be more efficient while still providing the required information. Suppose we have a huge
quantity of a yarn, say, yarn lakhs of bobbin 1000’s of ring bobbins are there. And if we want
to test all the bobbins that is total material for say it is count or it is strength.

It is not feasible. So we have to select the representative sample to evaluate any attributes.
And this if we can sample properly scientifically, this will give sufficient information and
which will be actually useful.
(Refer Slide Time: 02:35)
So, the question is that what does sampling mean? So it is a relatively small fraction of
material, which is selected from the respective population. And the selection is not random.
There are certain processes of selection so main reason of sampling, so why do we need
sampling? We can go for testing of whole material if I have one product, one item. I will go
for testing. Okay.

This is, I will test everything. I will try to test everything. But then why do we need
sampling? Sampling is required because we do not have a small quantity of sample. We have
huge quantity of sample, large quantity of sample. Population is a large quantity. So if we
want to test the tot entire population, then it requires time. And we do not have time. We need
the result immediately.

For example if you want to know the evenness of yarn say 30’s count yarn cotton yarn, which
has been produced in particular factory of say 1 lakhs spindle. So per day it is producing say
more than 1 lakhs bobbin this type of bobbin. And if you want to test the entire population
then it will be time consuming. So we cannot get result immediately. So we want result
immediately. So we cannot test the entire population.

So we have to take the small fraction of that. So time it is related with the cost. So when it
requires manpower it requires everything, the instrument the number of instrument, power,
everything, so it is cost is enhanced. Next reason is that most of the textile materials or testing
are the destructive in nature. So if we keep on destroying the population then we will not get
the material, actual material. So we have to select the sample.
So, these are the 2 basic reason for which we have to go for Sampling. And sampling is
required basically for textile material. It is very important. Because textile material it is
variable in nature. So if we see a particular product which is where variation is not there.
Like, one industry who is producing a ball particular ball, say iron ball, by say certain method
and that there is a fixed system, fixed process and it will keep on producing the ball of same
diameter, same dimension; a rod.

An industry is producing a rod. Okay. Iron rod. It will be having almost same dimension so
variability is not there. In that case sampling, all the sampling will be required but it is not
that critical. But in textile material main problem is that most of the textile material be it a
fibre, yarn, fabric, wherever it is a variable in nature. You take any characteristics. Diameter
it is a variable in diameter. Strength it varies from place to place. So sampling here it is
extremely important just one small quick example of sampling.
(Refer Slide Time: 07:17)

Suppose I want to a sample 100 gram of cotton from say 10 tons of cotton. So what is the
fraction you are trying to see? So
1
---------------
105
that small fraction we are taking for testing. Similarly say 10 random sample of cone we are
trying to take from 15 ton consignment. So we are taking a small fraction. So by doing these
things we are saving time saving cost. And we are saving the material also. But the sampling
has to be perfect. So that it gives confidence, that it is actually giving indicating the
characteristics of the population.
(Refer Slide Time: 08:22)

When the set of all possible items in a population is very large, it may be too costly or too
time consuming to do a comprehensive analysis of all of the items. So if you take another
example, for example ring bobbins in a spinning mill. It may be too expensive, to test all the
bobbins to determine the yarn characteristics, test strength.
(Refer Slide Time: 08:58)
So the basic aim of sampling is to produce an unbiased representative of whole population. It
should be free from biasness. So the result of sample may be different from the population.
But the sampling should be unbiased. There should not be any; its only very variation should
be in the chance but none other than that. For example in cotton fibre sampling from bale, the
proportion of fibre length in a sample should be almost same as the proportion in the bulk.

Then only we can test the sample for length or diameter. So if we do random sampling, so
unbiased sampling then we can tell with certain confidence the proportion whatever the
proportion of length whatever they are in the sample is almost the proportion there in the
bulk; in the bale. If we take the sample from the bale or from the bulk at random from
random places then we can take.

But if we have some biasness then we will not get this same proportion. Or through sampling
system each fibre in the bale should have equal chance of being chosen for sample. So you
are giving all the fibres equal chance for to get selected, means, if it is a fibre is in the bale
form, if the person, testing person testing goes he is picking the sample from the surface from
the outside the Bale.
(Refer Slide Time: 11:26)

Now let us see suppose this is a bale. Okay. Fibre are there different fibres are there. Now if
the system says that he will only pick the fibre from the surface. Then the things will be that
that if it is known, if it is the standard then what people will start doing the fibre manufacturer
okay. Fibre supplier they will do they will give smaller fibre at the centre covered with the;
with long fibre. Then we are getting wrong sample.
So the sampling is telling that there should be no biasness. We have to take the sample from
everywhere. The chance of sampling is there from everywhere. So the, all the fibre from the
centre of the bale, from the surface of the bale, from the bottom of the bale there are fibres,
they have the equal chance. Okay. And For that we have to select some sampling process.
(Refer Slide Time: 12:59)

The governing factor for the sampling is that that means what type of sampling we will
decide, we will adopt. It is decided by the form of the material like whether it is a fibre form,
it is a yarn form or it is a fabric form. Okay. If it is fibre, whether it is a loose fibre or it is in
bale form or the fibre is from the sliver so the sampling method of fibre, so we are taking say
fibre, this is the fibre and here also we are taking fibre.

They are the same. It is fibre. But if the form of material if it is in the loose form, the
sampling technique will be entirely different from the fibres if it is in the sliver form.
Similarly if we want to test the fibre characteristics from roving which is twisted? So here the
testing method will be totally different so sampling method. So sampling method is entirely
governed by the form of the material. Suppose we want to test the yarn, from the bobbin.

This is the sampling technique will be different. Then if we want to test yarn from a fabric, if
we want to test yarn say we want to test twist of yarn from fabric the sampling technique will
be entirely different. Next is the nature of material, nature of material of same form like, as
we have discussed the form whether it is in the form of bale, in the form of sliver, in the form
of yarn but if it is in the same form like bale. Bale in wool and bale in cotton, it is entirely
different.

Because cotton, if we open you can take out fibre because it is just compressed. But wool in
bale form it is entirely different because it contains huge quantity of grease which make it
very compact. Okay, so that fibre sampling from bale of wool is different from fibre sampling
of from the bale of cotton. So, from Bale of cotton we can pick the fibre, by hand. But for
wool we cannot do.

So for wool sampling, wool fibre sampling the system is different, that we will discuss. Okay,
the amount of material available. Suppose I have a material of a container of say 15 ton of
yarn. So 1000’s of bobbins are cones are available. In that case I will adopt a particular
sampling technique. But I have if I have only say, 10 bobbins are available. In that case the
sampling technique will be different .Okay.

Nature of test, so based on the nature of test, we have to adopt the different sampling
techniques like high volume instrument where you need sufficient quantity of material. Okay.
In that case you can actually take different sampling technique. Then if we need a single fibre
length measurement, in that case we have to adopt different sampling technique. Type of test
instrument, so whether we are going for a slow method or very fast method.

Or different types of same parameter, like again it is a HVI test or it is a comb sorter so
different types of test method, test instrument we have to sample differently. Information
required. For example say diameter. Do you want to measure diameter? Or we want to
measure fibre length. We will see in wool testing one sampling technique is called core
sampling technique. In core sampling technique where a coring tube is there.

Coring tube penetrate inside wool bale where which cuts the wool fibre. So which cuts and
take the core, take the sample from the inside the bale. Now once it is cutting the fibre that
means do you want to test, do you test length for that? We will not test. The coring technique
is used for either diameter measurement or may be grease content measurement. If you want
to test the length of wool fibre then it will be, sample technique will be entirely different.
Okay.
So what information do we require based on that requirement we change the sampling
technique? Like twist if we want to measure the twist the sampling techniques will be entirely
different from the; if we want to measure the lea strength. So total different sampling
technique will be there because for twist we need small, smaller length, for lea testing we
need say 100, 20 yards length. Okay. So that how we have to change our sampling technique.

Degree of accuracy required. So if we think the okay, we want to only know the
characteristics. Okay. Overall characteristics then our sampling our number of sample should
be will be less. We can have, say, from single bobbin we can have test just to know the get
the idea. But if we want to have certain confidence say 99% confidence level or 95%
confidence level, we can tell okay this is the range required. Okay. In that case we need a
large number of samples. So sampling techniques has to be changed.
(Refer Slide Time: 20:33)

Then the sampling type, the sampling type is broadly divided into 2 categories. One is called
Statistical sampling. And next is the Non statistical sampling. Statistical sampling means it is
the process, sampling techniques these are the techniques where you use bit of statistics.
Some randomness you use and Non statistical where we do not need statistical technique,
okay.
(Refer Slide Time: 21:10)
The statistical sampling again can be sub divided into 4 categories, 4 types. First is called
random sampling then systematic sampling. Third one is Stratified sampling which is based
on the strata means sub group. We are sub grouping that. Stratified sub grouping and 4th one
is the Cluster sampling. So and the non statistical sampling is subdivided into 2 categories.
One is called haphazard sampling and then judgemental sampling. So these are the different
types of sampling. We will start with the Statistical sampling and Statistical random
sampling.
(Refer Slide Time: 22:15)

So, random sampling is based on a random number generation. In this method each item in
the population has the same probability of being selected. As a part of sample has any other
sample material. So they have the equal opportunity of being selected okay. Just for example
a person testing person could randomly select 20 bobbins. So his idea is to select 20 bobbins.
Okay, to test case of population bobbin within the range of 1 to 1000.

So, 1000 bobbins are there he has to select the only 20 bobbins, okay, to execute some
testing. So, how to do that? What he will do? He will, he can use 2 techniques.
(Refer Slide Time: 23:35)

One is he can use random number. He can generate random number. So from 1 to 1000, 20
random numbers can be generated. And he picks the sample at 20 times. But here one has to
be very careful that the population is numbered. It is a random. There is no biasness in
numbering. So the lab technician could use a random number generator or he can simply, one
is random number generation. Or he can simply put the number 1, 2, 3, 4, 5 and 1000 number
he can put in small slip and randomly put take like random lottery.

And this 1 to 1000 are like 1000 spindles in ring frame. There are 1000 spindles in ring
frame. And he has generated the random number 20 numbers. And he picks the bobbin from
those all those spindles. And here the as it is randomly generated all the bobbins, they have
the equal opportunity. So that is why how he is using some statistical technique and it is
called random sampling.

And one advantage is that in random sampling he can use with or without replacement which
means suppose he has to test 5 times. And what he is doing, so with replacement means so
suppose he after random generation he has taken so 51st bobbin, then 92th bobbin and then
100 something bobbin like such 20 bobbins. Now with replacement means he will again
replace these bobbins to that place.

Particular their places he will replace. And again he will draw the random number. That
means the same bobbin has got chance of being selected there is no issue because he is using
random number. So random sampling can be done with or without replacement. And if he
feels this I will not take back all, this bobbin. So that means on 20 numbers whatever he has
generated, he will discard. He will next sample, he will take the replacement sample. He will
take the other samples. So this is done.

So, if it is done without replacement and item is not returned to the population again, after it
selected okay and thus only it can occur only once. Okay. So that is why. But he can it is up
to the system he can do with or without replacement. After random sampling next is the
Systematic sampling.
(Refer Slide Time: 27:11)

Here it is another statistical Method, here the system that one has to be careful here. In this
method every nth sample, nth element he will take from the list of selected. Okay. Starting
with the sample element n randomly selected from the k element. So there are k elements,
every k element, so he will pick nth term. I will give example. If the population has 1000
bobbin, so, here it has got 1000 bobbin and ultimately we have to select, our idea is to select
100 bobbins. 100 bobbins are selected. That means if you divide 1000 / 100, K’s value is 10.
So, everything 10th segments every 10th subgroup he has to pick 1 sample. So here K value
is a 10. Among 10th suppose he decides okay and among every 10 bobbin, I will select 6th
value 6th bobbin. So, if number 6 which is n is randomly selected from first 10 K value first
10 elements on the list. The sample would continue down the list selecting the every 6th
element from each group of 10th element that means what? If it is 1 to 1000 He will start
taking 6th, 16th and 26th like that. So this nth term it is actually one can decide randomly.
Okay.
(Refer Slide Time: 29:24)

So, care must be taken when using the systematic sample to ensure that the original
population list has not been ordered. So it is a population list has to be random. Okay, order
in any way which actually can introduce non random list. So the original population has to be
totally random. There is no biasness similar example one can give here. It is a bobbin of 1000
spindles. So every 6th, 16th, 26th, and bobbin he can take.

Another example you can see same, in spinning mill if the auditor of the buying house selects
the 14th ring bobbin out of 20th, in a random list of all the bobbins. It is randomly here so
arranged here. So, 14th he is trying to take. Then what are the numbers you will take? First
you will take 14th, and then you will take 34th, then 54th, 74th, in this way. So after
Systematic sampling,
(Refer Slide Time: 30:46)
Next is that stratified sampling, which is actually this is the statistical sampling. This
sampling is used, when representative from each subgroup is known as strata within the
population need to be represented in the sample. It is not like a random sample. Random
sampling it may or may not incorporate a particular group. Like in an industry who is
producing 30s count yarn.

From same shed, ring frame shed, it has got a 100 ring frame. So it is producing 30’s count.
And the ring frame they are from different manufacturer, manufacturer 1, manufacturer 2, 3,
4, 5 manufacture. Now random sampling says that out of 1000 spindle you take random
depending on the random number generation. Do not concentrate on the manufacturer.

Manufacturer A, B whatever, you can take if it is coming random number, if you may select
the 10 bobbins from the same manufacturing. But stratified sampling says it is not. You first
sub group the make. Sub group the manufacturer first. So manufacturer 1, manufacturer 2,
manufacturer 3, manufacturer 4 and then decide how to take the sample. And you have to
take sample from each and every sub group.

Either you decide, I will select say 4 sub groups are there and we have to take say 20 bobbins.
So, I will, deciding I will decide to take 5 from each manufacturer. And this 5 when you have
selected, decided then you go for random number. But here in stratified sampling the
advantage is that your population their samples represent the all the make all the
manufacturer.
The first step is to divide the population in to subgroups. First you have to divide based on
mutually exclusive criteria. So, you have to actually decide, you have to mutually understand,
mutually exclusive criteria. Like the, make of ring frame in spinning mill producing a same
yarn. Now suppose I am deciding okay. This will be equally that I have given example. But
one can decide in other way I will give one example. And then random or systematic sample
are taken from the each sub group.
(Refer Slide Time: 34:20)

Now example is that if 40 ring bobbins are to be selected, so I want to select 40 ring bobbins
where 10 percent of ring frames are make of 1, 60% are of make 2, 25% are of make 3 and a
5% of make 4, 40 you have to take. Now the mutually understand criteria is that we have to
select the bobbins, number of non-based on the proportion of ring frame.

The large proportion we will have larger quantity in the sample. So then in that case 4
bobbins from make 1 because it is a 10% of the ring frame. So 4 bobbins we will take out of
40. As it is 60% of make 2 so then 24 bobbins I will take from there. Okay. Then 10 bobbins
from make 3 and 2 bobbins from make 4. So it takes it take care of the 40 bobbins. But we
may fix other criteria. Okay.

All, the make should have equal representation. So 4 make10 from each make whatever
maybe the number I will take 10 from each so that it is equal. So it is mutually agreeable
criteria. And then once we decide then we have to do random selection. Like 4 bobbins from
make 1 this 4 number of bobbins I may use random number. Then we will take 24 bobbins
from make 2. I will use random number to get 24. But it is okay. Stratified samplings can also
sample and equal number of items from each sample as we have mentioned okay.
(Refer Slide Time: 36:24)

Next is that that it is a Cluster sampling. It is also known as the block sampling. It is different
from the stratified sampling. In Cluster sampling, the population that is being sampled is
divided into group called Cluster. That means instead of these sub groups being
homogeneous based on a selected criteria as in stratified sampling. The cluster sampling is
heterogeneous, as heterogeneous as possible, okay to match the population.

Like there is same make say 100 ring frames, we make a cluster. Okay. We will test, today
we will test from this cluster, take testing okay sample, then and it is next day we can go for
in other cluster. And a random sample is then taken like your stratified sampling once we
have clustered, we have blocked the sub group then we can take the random sample.
(Refer Slide Time: 37:51)
Just for example if a spinning mill has 30 ring frames running, the sampling person looking
for testing of yarn. Linear density might use a cluster sampling to randomly selected 4 of
those ring frames as representative. Okay only 4 we have selected. And then he is taking
cluster. He is not taking from all the ring frames for the test and then randomly selects
samples from them. This is actually used in, spinning industries also. This type, he is not
taking the entire. Cluster sampling is very popular very commonly used here okay.
(Refer Slide Time: 38:36)

Cluster sampling can tell us a lot about the particular cluster. That is problem here. But unless
the clusters are selected randomly, so we have to again select the cluster randomly. Then if it
is not done, then it will give us wrong result. Okay. So generalisation cannot always be made
about the entire population. So Cluster sampling is used in various ways. Like a different
shift people try to that is, it is a cluster. So at night shift he is trying to take the sample from
the night shift production, okay.

He is taking and he is testing that may result in entirely different result than the population.
So Cluster sampling, they have got draw backs. So random sampling from only 4 ring frames,
or ring frames from particular make or from product of a particular shift may cause biases to
the sample. That would not allow statistically valid generalization so one cannot generalize. It
is a simple one. Okay. It is a feasible one.

But it you cannot generalize. It is not possible some time to test suppose testing person who is
working in night shift he will only test the material at night shift. But it gives certain result; it
is not giving the population, total population. So that particular cluster particular shift it is
giving idea. It is giving that idea of that test results. Say strength. But if you want to know the
population then he has to mix all the shifts. Then Cluster sampling is not going to be useful
okay.
(Refer Slide Time: 40:37)

And in Non-statistical sampling there are 2 types of Non-statistical sampling we have

discussed. In Non-statistical sampling the information about the entire population cannot be
extrapolated from the sample. So we get some information which is quick enough, which is
easy very quick and sometime it is useful. But you cannot extrapolate, you cannot predict to
the entire population.
Suppose what I am doing I am picking one bobbin from the entire population. I am just
picking a bobbin. I will get some information okay. This lot is s giving yarn strength or like
that. But I cannot tell with certain confidence that Entire population is like that. Because I
have not I have done in the statistical method. Non-statistical samplings are of 2 types as I
have mentioned Haphazard sampling and Judgemental sampling.
(Refer Slide Time: 41:44)

In Haphazard sampling the samples are selected based on convenience. But preferably should
still be chosen as randomly as possible. The best and convenience means the tester of person
is going and he picks our bobbin from the top. This is in textile industry it is a very
commonly used very common. Like, senior management person, President or some chairman
when he is moving, he would like to know the characteristics of the material.

He may ask the person you pick that bobbin. And test and let me know the result, okay. So
that will give us the quick idea about the product okay. But we cannot come into any
conclusion that because it is not statistically sampled. The example is that are testing person
may go to the shop floor and then close his eyes and point at a box containing the ring frame
to select a bobbin just You take this one. That is the haphazard sampling.

That is basically here it is not time consuming. He is going there and picking and
immediately you will give the result. He could also grab the bobbin close to him, from the
bottom or from top anywhere so that is the haphazard sampling.
(Refer Slide Time: 43:37)
The advantage is that it is quicker and use a smaller sample size than other sampling
technique. So you can take, pick okay 2 bobbins and test it so that is the haphazard sampling.
Advantage of haphazard sampling and the main disadvantage is that since it is not statistically
based generalization we cannot do about the total population should be made with extreme
caution. That is very important we may tell okay this is the strength. This is the evenness but
we cannot generalize and until and unless we are doing some statistical sampling. It is only
for idea
(Refer Slide Time: 44:30)

Next is that it is a judgmental sampling. In the judgmental sampling which is again non-
statistical a person doing the sampling uses his or her knowledge or experience in selecting
sample. It is judgemental. Here suppose a particular in a ring frame shed, a particular ring
frame gives problem always. It is supposed may be due to some humidity problem in that
particular portion or maybe due to vibration.

It generates high unevenness, high U% because of the vibration in spindle. And after trial
after various your maintenance this could not be rectified. So the problem would be that then
what he will do he will try to avoid that sample. He will try to avoid taking sample from that
particular ring frame. That is he is using his judgement so that his sample his total entire
population does not get affected.

It is based on his experience. The example is based on his experience. The person may know
which type of item have more chance okay, to have non-conformance, what type of item?
Like product of night shift suppose sometimes it creates a problem. Or some product of the
see first few lot, first few doff in the ring frame he is not taking, based on his experience
because it will get non-conformance.

Or which type of item had problem in past. This type of item had problem okay I will not
take, I will not take risk of the rejection. I will not take any risk. So I will justify like he will
try to take the sample which is a bobbin which is very good looking. Not he is not going to
take damaged bobbin he is not he is trying to take fresh look sample. So that type of sample
he is using his judgement, okay.

So that it is not wrong but he is rejecting some non-conformance basically. He is not trying to
take any risk. That is the he is using the judgement. Another example is that the sample tester
might select bobbin with better appearance he always tries to take. Or other sample selected
he would like to take okay only I will take the problematic one. This is judgement okay. And
his judgement sampling is important it is happens in actually buying house, in actual trade.

Like the supplier will always try to use judgmental sampling based on the best quality. He
should not take non-conformance. But the customer who will try to see the where is the
faulty, where is the fault. Suppose you are sending a consignment to a supplier. He is telling,
okay you sample yarns are rejected. It is giving a problem. What is the problem your samples
are there, bobbins are damaged.
He will send say 5 to 10 bobbins are package as damage package. Once you are visiting to his
place it is everything is going alright because he is using his judgement wrong judgement.
Okay. So it is actually sometime it creates biasness.
(Refer Slide Time: 48:57)

So another way of broad classification of sampling is that one is it is particularly for textile
material. Earlier sampling techniques were for general sampling technique and this technique
is for specifically for textile material. One is called numerical sampling. Here in this
numerical sampling, a sample in which proportion of all the Fibres in sample is exactly equal
to the proportion of the population. So that means it is strictly as per set norms.

We have to take that sampling that, which means if you take out if you remove one sample,
the rest population, the proportion of fibre length will to get affected. That is called the
numerical sampling. I will keep example subsequently, okay. And next is the biased
sampling. Biased sampling it is a basically in textile, in a fibre, it is called length Biased
sampling.

In Biased sampling when the selection is influenced by factors other than chance that is
called biased sampling, a sample ceases to be truly representative of the bulk and a biased
sample result. So biased sample is actually where it is not due to the chance factor variability
is due to something else okay. Now let us see what are the different causes of biasness?
(Refer Slide Time: 50:49)
Biasness can happen due to various reasons the causes of biasness is that bias due to physical
characteristics. As I have already mentioned the longer fibre will have always greater chance
to be selected because it has got higher surface area. So in this sliver it has got long fibre or
short fibres mixed. But if I am trying to pick the fibres higher probability will be there that
longer fibres are selected, because it is a higher area. Okay higher projection.

Next is that a cause of biasness is that position related to the person who is selecting, he is
taking the bias sample. The lab assistant may pick bobbin from the top or may pick bobbin
from which is close to him. So that is one cause of biasness. Third cause is that subconscious
biasness that I have already mentioned which is called haphazard or this type of sampling,
subconscious biasness, we have tendency to pick best looking item and free from any
damage.

That type of sample we are taking. That means we are introducing biasness. We are trying to
get the best result that is the sample. And also as I have already mentioned sometime it is a
conscious biasness is there. As a supplier you are trying to project your product is a best
quality. As a customer I am trying to project that your material which is worst. So that type of
conscious biasness is also purchased. But it is not that important.

These are the causes of biasness. Next class we will continue with this sampling, okay. Till
then okay goodbye.