8/9/2019 Sampling & Sampling Distribution
1/36
8/9/2019 Sampling & Sampling Distribution
2/36
Objectives
To beconversant withtheterms The advantages of sampling Types of Sampling Sampling Distributions Sampling Errors To learn Central LimitTheorem
8/9/2019 Sampling & Sampling Distribution
3/36
Why Samplethe Population?
Physical impossibility of checkingall items (census).
Thecost of studying all the items.
Highlytime-consuming.
8/9/2019 Sampling & Sampling Distribution
4/36
The destructive nature of certain
tests.
Thesample results areusually
adequate.
8/9/2019 Sampling & Sampling Distribution
5/36
Probability Sampling
Sampleselected suchthateachitem or person in the populationbeing studied has an equal
likelihood of being selected
8/9/2019 Sampling & Sampling Distribution
6/36
Methods of Probability Sampling
Simple Random Sample:A sample so formulated that each itemor person in the whole population has thesame chance of being selectedeg: lottery , housie.
Systematic Random Sampling:
The items or individuals of the populationarranged in some order.Arandom startingpoint is selected and then every kth memberof the population is selected for the sample.
8/9/2019 Sampling & Sampling Distribution
7/36
Stratified Random Sampling:Heterogeneous population first divided intohomogeneous subgroups, or strata,having
somecommon features.Proportionate subsample isselected fromeachsuchstratum at random and combined toget a truesample.
Cluster Sampling:
A population first divided into primaryunits orclusters, which is asheterogeneous asthepopulation .Then random samples of clustersselected for thesurvey.
8/9/2019 Sampling & Sampling Distribution
8/36
Non Probability Sample
Inclusion in thesamplebased on thejudgment or convenience.
Many items may not get a chance.
Statistical theoriescannotbeused.
8/9/2019 Sampling & Sampling Distribution
9/36
Sampling Error
The differencebetween the value of a samplestatistic and the value of thecorrespondingpopulation parameter .
In thecase of mean, Sampling error = X QIn case of Std deviation,
Sampling error = s- W
Assuming thatthesample is random and nonon sampling error hasbeen made
8/9/2019 Sampling & Sampling Distribution
10/36
Non Sampling Error
Errors dueto deficiencies in thecollection , recording and
tabulation of data.
8/9/2019 Sampling & Sampling Distribution
11/36
Sampling DistributionMany number of samples of size n can be drawnfrom a population of N (N C n).
Highly improbablethat all samples havethe
same mean.They vary from sampleto sample.
The probability distribution of all these values ofmeans of varioussamplestaken from a populationiscalled Sampling distribution of means.
Sampling distribution of std deviations,proportionsetccan also be made.
8/9/2019 Sampling & Sampling Distribution
12/36
The law firm of Hoya and Associates
has five partners.At their weeklypartners meeting each reported thenumber of hours they worked fortheir clients in the last week.
P a r tn e r H o u r s
1 . D u n n 2 2
2 . H a r d y 2 6
3 . K ie r s 3 0
4 . M a l in o w s k i 2 6
5 . T i l lm a n 2 2
8/9/2019 Sampling & Sampling Distribution
13/36
If two partners areselectedrandomly,how many differentsamples are possible?
This isthecombination of 5objectstaken 2 at a time.That is:There are a total of 10 different
samples.
10)!25(!2
!525
!
!C
8/9/2019 Sampling & Sampling Distribution
14/36
P a r t n e r s T o t a l e a n1 , 2 4 8 2 4
1 , 3 5 2 2 6
1 , 4 4 8 2 4
1 , 5 4 4 2 2
2 , 3 5 6 2 8
2, 4 52 26
2, 5 48 24
3, 4 56 28
3, 5 52 26
4, 5 48 24
8/9/2019 Sampling & Sampling Distribution
15/36
Organizethesample means into asampling distribution.
S a m p l e
e a n
F r e q u e n c y R e la t i v e
F r e q u e n c y p r o b a b i l i t y
2 2 1 1 / 1 0
2 4 4 4 / 1 0
2 6 3 3 / 1 0
2 8 2 2 / 1 0
8/9/2019 Sampling & Sampling Distribution
16/36
Computethe mean of thesamplemeans. Compare it withthepopulation mean.
2.2510
)2(28)3(26)4(24)1(22!
!
XQ
8/9/2019 Sampling & Sampling Distribution
17/36
The population mean is also 25.2 hours.
2.255
2226302622!
!Q
Notethatthe mean of thesample means isexactlyequal to the population mean if allpossible samples are taken. But not otherwise.
8/9/2019 Sampling & Sampling Distribution
18/36
Standard Error Standard Error (S.E) of a statistic isthestd deviation of sampling distribution ofthatstatistic.
S.E of Mean , Wx = W /root of n S.E of Std Deviation,Ws = W /root of 2n
S.E of proportion,W p = root of (pq/n)
Std error isused to test if the differencebetween samplestatistic and populationparameter issignificant or not.
8/9/2019 Sampling & Sampling Distribution
19/36
A random sample of 500 oranges wastaken from a largeconsignment and 65were found to be defective. Find theStandard error of the proportion of bad
ones in a sample of thissize.
p= 65/500 =0.13
q= 1-p = 0.87, n = 500 Std Error = root (pq/n) = 0.015
8/9/2019 Sampling & Sampling Distribution
20/36
8/9/2019 Sampling & Sampling Distribution
21/36
A simple random sample of size 36 is
drawn from a finite population consistingof 101 units.If the population Standarddeviation is 12.6 find thestandard error
of sample mean when thesample is drawni. with replacement ii. withoutreplacement
Ans: i.W/root n =2.1,
ii. =(W/root n) * root {(N-n)/(N-1)}
= 1.693
8/9/2019 Sampling & Sampling Distribution
22/36
Central Limit Theorem
Thesampling distribution of the means ofn samples generated from a populationwill tend towards normal distribution, as
the value of n increases.
This is irrespective of the distribution ofthe population itself.
The mean of thesampling distribution ofmeans will equal the population mean.
8/9/2019 Sampling & Sampling Distribution
23/36
Normal Uniform Skewed
Population
n = 2
n = 30
XQXQXQXQ
General
The Central LimitTheorem AppliestoSampling Distributions from AnyAny Population
8/9/2019 Sampling & Sampling Distribution
24/36
The mean of thesampling distributionof thesample means, X tendsto Qand itsstandard deviation Wx tends toW / root n,
For largesamples, (n > 30),
X can betaken as = Q
And Wx as = W / root n.
8/9/2019 Sampling & Sampling Distribution
25/36
Purpose of sampling The main purpose of sampling isto infer the
parameter from thestatistic.eg:exit poll
Itcan beused to identifytheunknownpopulation from which a given samplecan beexpected to belong.eg :Oil spill, blood stain
Itcan beused to assesthe probability of aone or moresamples belonging to thesamepopulation.eg:twin sisters
Central LimitTheorem permitsusto usenormal distribution to testthese, if thesample size is largeenough.
8/9/2019 Sampling & Sampling Distribution
26/36
Example 1
The weight of persons using a lift isnormally distributed with a mean of 70 kg
and a standard deviation of 9 kg. The lifthas a maximum capacity of 300 kg.Iffour persons enter the lift, what is theprobability of the load exceeding the
limit? Hint similar to marks / first class case
Sample mean how far away from mean of mean
8/9/2019 Sampling & Sampling Distribution
27/36
Population mean Q = 70 kg
Sample mean = 300/4 = 75 kg
Thesampling error of sampling distribution
of means from a population is = W/root n
= 9/root 4 = 4.5kg
P( x> 75 ) isto be found out.
z = x-Q/Wx = (75-70)/4.5 = 1.11 From normal table,corresponding
P (z >1.11) = 0.5 -.3665 = 0.1335
8/9/2019 Sampling & Sampling Distribution
28/36
Example 2
From past experience it is knownthat the std deviation of thenumber of days of absence ofworkers in an industry is 15.In arandom sample of 100 workers, whatis the probability for the mean
value to differ from the actual meanby more than 3 days ?
8/9/2019 Sampling & Sampling Distribution
29/36
Using normal distribution,
Standard error of sampling distribution
Wx = W/root n =15/ root
100 =
1.53 days = 2 Wx
>3 days (ie x) = >2 (ie z) from the mean
Hence P (x > 3 or < -3) = P (z>2 or < -2).
From table, this is =(.5-.4772)+(.5-.4772)=0.0456
8/9/2019 Sampling & Sampling Distribution
30/36
8/9/2019 Sampling & Sampling Distribution
31/36
Example 3
In a normal distribution with mean 375 andstd deviation 48, how big a sample shouldbe taken so that the probability for the
sample mean to fall between 370 and 380will be at least 0.95? Ans :
Probability on either side of mean = 0.475.
From table, corresponding value of z =1.96
1.96 = (380-375)/(48/root n) . n =354.3
So sample size to be at least 355.
8/9/2019 Sampling & Sampling Distribution
32/36
An oil refinery has back up monitors to keeptrack of the oil flow to prevent disruption inprocess.A monitor has an average life of4300 hrs and a std deviation of 730 hrs.Ithas two standby identical units which willautomatically come into operation, if anymonitor fails. What is the probability that aset of monitors will last for
i.13000 hrs at least? ii.12630hrs at the most?
Hint: Treat it as a sample 3 monitors.
Ans: i. 0.4685 ii. 0.4154
Example 4 / HW
8/9/2019 Sampling & Sampling Distribution
33/36
Treat it as a sample 3 monitors. Average life of population Q = 4300hrs. W = 730 hrs. S.E = 730/root 3 = 421.5i.To get a life of 13000 hrs with 3 monitors
working one after other,the mean of sampleshould be 13000/3 hrs = 4333.33 hrs.z = (4333.33 - 4300)/421.5 = 0.079
Probability (z > 0.079) = 0.5-.0315 = 0.4685ii. z = (4210 4300)/421.5 = - 0.2135Probability (z< -0.2135)= 0.5-0.0846= 0.4154
8/9/2019 Sampling & Sampling Distribution
34/36
Example HW
An underwater salvageteam is planning toexplorethesite where 45 Spanishshipscarrying gold sank. From records, it appears
that itcan generate a revenue of 225,000$per ship, and thestd deviation isexpected tobe 39,000$.Theteams financier isskeptic andopined that if theexpedition expenses of 2.1
m$ is not recovered in the first 9 ships,he willcancel the reminder of theexploration.Whatisthe probability for theexploration tocontinuebeyond 9 ships? Ans : 0.2393
8/9/2019 Sampling & Sampling Distribution
35/36
Sara Gorden is. 6.46 LEVIN
Example HW
8/9/2019 Sampling & Sampling Distribution
36/36
Assignment- Sampling
Sampling-page 311-6.3,19,21 Levin Page 323- 6.39,
6-46, 6-54, 6-62,6-64
Page 339 for objective