Topic 7

Sample Selection

HOTL3003

Research for Business and Tourism

Readings

⚫Study Guide Topic 7

⚫Textbook: Quinlan et al. (2015)

Business Research Methods Chapter 10

pp. 168-189.

Research as a process

The research process Quinlan et al. (2015, p. 3).

The four frameworks approach

Model of the four frameworks. Quinlan et al. (2015, p 7).

This Week’s Objectives

1. Discuss the role of sampling in social science research

2. Define sampling terminology

3. Distinguish probability and non-probability

sampling techniques

4. Discuss the range of probability and nonprobability sampling techniques and the

situations in which they are applied

5. Outline factors to consider in determining the

necessary sample size

6. Recognise types of sampling errors

What is sampling and why do we need it?

⚫The practice of studying ALL the units or cases in a

population is known as a census

⚫However, this is normally not possible because the

population may be too large to survey everyone.

⚫Studying a sample allows research to be conducted using

significantly less resources than studying entire

populations

⚫Sampling is about selecting a small number of units or

cases from a larger collection or ‘population’ for

inclusion in a study

Why use a sample?

⚫Studying all these passengers would be prohibitively expensive and time consuming

The purpose of this study is to determine passenger satisfaction levels with inflight service provided by the Australian low-cost airlines Virgin and

Jetstar between 1st July and 31st December 2016

⚫How many passengers would fly with Virgin and Jetstar during this period?

⚫Tens of thousands, hundreds of thousands, millions…?

Australian domestic air travel 2015-16

https://bitre.gov.au/statistics/aviation/d

omestic.aspx

BUT…

⚫We could design our study so that we randomly

select a sample of say, 3000 passengers

⚫This would drastically reduce the resources needed

to conduct the study, and

⚫So long as we select our sample based upon accepted

probability sampling methods, the results would

be representative of that wider population of

passengers

⚫This means that studying a sample is effectively as

good as surveying every person that has flown

with these two low-cost airlines during that period

Some sampling terms you need to know

Quinlan et al. (2015, p. 177)

Sampling Terminology

Population/Target population

⚫A complete group (of persons, things, events – called

units/subjects/elements) sharing some common

characteristics

⚫ALL the units that are the focus of a particular research

project

⚫E.g. every single passenger who has flown with Virgin or

Jetstar between 1st July and 31st December 2016

Census

⚫Investigation of all the members of a population

⚫E.g. we study every single passenger who has f lown with

Virgin or Jetstar between 1st July and 31st December 2016

Sampling Terminology

Sample

⚫A subset of a larger population.

Sampling

⚫The means by which units are selected for

inclusion in the sample

⚫E.g. every 5th passenger on manifest lists on each

flight during the study period

Sampling Terminology

Sampling Ratio

⚫The ratio of the sample size to the size of the target

population

E.g.

Target population = 400,000

Sample size = 3,000

⚫Sampling Ratio = (3,000/400,000) x 100 = 0.75%

⚫This is the percentage of the population we

intend to study

Sampling Terminology

Sampling Frame

⚫A list representing all units of a target

population from which the researcher selects a

sample

⚫E.g. Virgin & Jetstar passenger lists from July to

December 2016

⚫Other examples might be phone books,

electoral rolls, list of ticket holders at an

event

Sampling Terminology

Sample Unit

⚫An individual unit selected from the sampling

frame for inclusion in the sample group

⚫E.g. from our airline example…

⚫A sampling unit = A passenger who flew with Virgin

or Jetstar between 1st July and 31st December 2016

What is a representative sample?

⚫A sample that accurately reflects the population from which

it was drawn.

⚫If the researcher designs the sample carefully s/he can

obtain a representative sample, which is one that allows

the researcher to produce accurate generalisations about

the larger group.

⚫This means that the researcher can say that the outcomes of

a study are true of the wider population (within a specific

margin for error) from which the sample was drawn – even

though only a small percentage of the entire population has

been studied

⚫Based on mathematical theory that a random sample can be

representative of a largergroup

⚫The word ‘random’ does not imply that the sample is selected

aimlessly but that each element has an equal chance of being

selected.

⚫E.g. If I were to close my eyes and walk around the room and

tap 10 people on the head to participate in a study, that is

not a random technique – it is haphazard as I have no

measures in place to ensure that each person has an equal

chance of being included in the sample

⚫Techniques include: simple random, systematic, stratified

random and multistage cluster sampling

Probability or Random Sampling

Probability Sampling Techniques

Simple Random Sampling

⚫Ensures ALL elements in a

population have an equal

chance of being included in

the sample

⚫Drawing names/numbers from a

hat

⚫For larger samples a random

numbers table is commonly used,

or the sample is computergenerated

61287 | 45055 | 49679 | 36251 |

58461 | 9182 | 73095 | 17734 |

49588 | 20104 | 74215 | 48480 |

95603 | 95559 | 2592 | 90874 |

76129 | 78145 | 34527 | 26170 |

87525 | 51118 | 53182 | 4321 |

39551 | 89221 | 3827 | 5851 |

53170 | 91435 | 77786 | 81074 |

22944 | 45992 | 54389 | 37900 |

39512 | 12888 | 84079 | 22020 |

54074 | 72689 | 38725 | 97027 |

46344 | 65052 | 65548 | 87960 |

77591 | 89487 | 84384 | 97329 |

83577 | 72528 | 60096 | 23866 |

87656 | 30817 | 27855 | 95997 |

23807 | 19873 | 51093 | 15889 |

13681 | 43462 | 11391 | 19290 |

Probability Sampling Techniques

Systematic Sampling

⚫Selects units / subjects based on

a pre- determined interval

⚫E.g. every 5th name on a

passenger list

⚫Often used in street / intercept

surveys e.g. every fifth

house/passerby, to obtain a

random sample

4 5 10

Probability Sampling Techniques

Stratified Sampling

⚫The researcherdivides the target

population into ‘sub-populations’

(strata) based upon certain criteria (e.g.

age, gender, location) on which we want to

ensure correct representation in the

sample.

⚫A random sample is then drawn from

each sub-population using simple or

systematic sampling.

⚫Undertaken to reduce sampling error

and ensure that each subpopulation is proportionally

represented in the final sample.

Passengers

100%

Business

30%

Leisure

70%

Probability Sampling Techniques

Clustersampling

⚫Useful when no sampling frame is available (and constructing one is

too costly) or when the population is geographically dispersed yet we

want to visit the sample in person.

⚫Involves drawing several different samples in order to obtain a final

sample.

⚫Start by drawing a sample of clusters of elements. Then for each

cluster, draw a sample of anotherclusteror individual elements. If

the former, then draw another sample until individual elements are

chosen.

⚫Example: We want to study cafes in Australian cities. Using

multistage cluster sampling we might firstly draw a random sample

of cities (clusters). Then, for each city in the sample we choose a

random sample of cafes (population elements).

Non-Probability or Non-Random Sampling

⚫In many instances it is not possible, practical or necessary to obtain a

random sample

⚫A suitable sampling frame may be unavailable or too expensive to

generate (e.g. all people have evervisited a particular town)

⚫Some research does not require the results to be generalised to the

wider

population, e.g. exploratory designs, pre-testing

⚫In these cases a non-probability or non-random sampling approach

is taken

Such samples do not accurately represent the population of

interest

⚫This is because subjects are not being selected randomly – i.e. they do

not have an equal chance of being selected in the sample

Non-Probability Techniques

Convenience Sampling

⚫Haphazardly selecting a sample or units / subjects because

they are easily accessible

⚫E.g. patrolling the baggage collection area in a Virgin / Jetstar

terminal and

selecting people who look friendly or approachable

⚫Should be used as a last resort – capable of producing

ineffective and highly unrepresentative samples.

Non-Probability Techniques

Purposive or Judgment Sampling

⚫The researcher uses her/his judgment to deliberately select the

most appropriate person(s) for inclusion in the study

⚫Participants are selected based upon their ability to provide the most

pertinent information

⚫Used most commonly in qualitative studies

⚫This method is often used in order to select difficult-to-reach,

expert, deviant, extreme, marginalized, misunderstood or

unique populations.

⚫E.g. purposively selecting human resource managers (as experts) in

an exploratory study of performance management of hotel

employees

⚫E.g. There is no list out there that identifies all problem gamblers

from which we can randomly sample. We can potentially recruit

people from counselling services to compile a ‘sample’ of problem

gamblers

Non-Probability Techniques

Snowball Sampling

⚫Another common sampling approach in

qualitative

studies

⚫Builds a sample through referrals.

⚫Used when members of a sample can only be

identified through their links with other

subjects. This often occurs with populations

that are not easily identified or accessed (e.g.

homeless people)

⚫E.g. you interview someone and they tell you ‘my

mate Joe Smith could tell you more about this’

⚫You contact Mr Smith for an interview and he

might

recommend another friend – the snowball

grows

Non-Probability Techniques

Quota Sampling

⚫Is one (small) step up from pure convenience sampling

as the researcher attempts to introduce a degree of

representativeness to the sample by ensuring different

groups are included

⚫Similar idea to stratified sampling but units within strata

are not chosen randomly, just chosen by convenience until

the required number of units has been met.

Once we have decided upon our

sampling technique, how do we know

how big our sample should be?

Sample Size – Quantitative Research

⚫Most quantitative studies aim to obtain large sample sizes

where possible

⚫To gain a higher degree of accuracy

⚫To facilitate in-depth analysis of sub-groups within the

sample

⚫A census is the preferred option – but this is rarely possible

⚫If a census is not possible, the sample size is often

determined relative to the size of the target population

and taking into account the homogeneity of the population.

⚫Also need to consider cost, time, data analysis methods

Sample Size

⚫There is no golden rule regarding how big a sample

should be

‘A large sample without random sampling or with a

poor sampling frame is less representative than a

smaller one with random sampling and an excellent

sampling frame’

(Neuman, 2003, pp. 231-232)

Neuman, WL (2003) Social Research Methods: Qualitative and Quantitative Approaches, Allyn and Bacon, New York.

Sample Size – Quantitative Research

⚫Increasing sample size increases accuracy but at a declining rate.

⚫Neuman (2003) argues that sample size should be a function relevant

to the size of the target population

⚫Populations <1,000, sample size

= ~30% (Population 1000,

sample 300)

⚫Populations ~10,000, sample size = ~10%

(Population 10,000, sample 1000)

⚫Populations >150,000, sample

size = ~1% (Population 150,000,

sample 1500)

⚫Also, the more homogenous the sample (i.e. most people are expected

toanswer in a particularway), the smaller the required sample size.

Sample Size – Quantitative Research

⚫When we speak about sample sizes in quantitative research we are

referring to the numberof returned, usable responses

E.g. if our target population is 6,000 and we post out 600

surveys (10%) – then we have got it wrong!

We really needed around 600 responses

So we need to actually send out a lot more than 600 surveys to

allow for non- response

⚫Sample size and response rates are two different things

⚫We aren’tconcerned with how many went out – we want to

know about how many came back

⚫All surveys will experience a degree of non-response – some

small, others extremely large to the point where data analysis

is not possible

⚫Pitfalls of postal surveys – low (20% average) response rate

Sample Size – Qualitative Research

⚫It is very difficult to plan how much data you will need to

collect in a

qualitative study

⚫However, one point of guidance is the following: Data is

collected until the saturation point is reached

⚫This is the point where no new information is

gathered/added to your study – it is said that sufficient

data has been collected

⚫Hence, when writing a research proposal the researcher

does not pre- empt how much data will need to be

collected – they state that they will continue until this

saturation point

Errors Associated with Sampling

⚫Sampling frame error: Sampling frame may not fully

represent population.

⚫Random sampling error: the extent to which the

sample differs from the population due to chance

variations.

⚫Non-response error: differences due to not all

selected sample units responding. Also called nonresponse bias.

⚫Plan to minimise these errors.

The steps in selecting a sample

1. Define the target population (draws on problem statement,

including scope)

2. Select (or compile) a suitable sampling frame (complete as

possible)

3. Determine whether probability or non-probability sampling

is required

4. Determine which sampling technique to use.

5. Determine the required sample size

6. Select sample units

Sampling & Your Assignment 4

⚫In the sampling section you must:

⚫State whether you are using probability or non-probability

sampling

⚫You should explain explicitly why yourselected sampling

technique is suitable foryour study

⚫ Good idea to include a paragraph on the sampling procedures that

could be applied to your study – noting their advantages and

disadvantages

⚫ Then narrow down to the one you have selected and justify your

choice with a reference or two. These references will most likely

be to research textbooks.

⚫ E.g. Neuman (2003) stated that stratified sampling was

appropriate under such and such circumstances, therefore

stratified sampling has been selected for the present study…

Sampling and Your Assignment 4

⚫Specifically, you should cover the following:

⚫Define your target population (draws on problem statement, including

scope)

⚫Whether probability or non-probability sampling will be used and why.

Link the “why” to the research problem, the availability of a sampling

frame and the research design.

⚫What sampling technique (e.g. simple random, stratified, purposive

etc.) will be used and why.

⚫What the planned sample size will be and why.

⚫Integrate references. In the sampling section, these would mostly be

references to research books when you are outlining advantages and

disadvantages and/or justifying choice of available/suitable methods.

⚫You might also come across journal papers in your area that also assist with

sampling method (e.g. “This research will follow the sampling approach

used by Smith et al. (2010) in a study of the same concepts but in a

different context…”)

Questions??

THANK YOU

hotelschool.scu.edu.au