Data Collection

RESEARCH METHODOLOGY MODULE 3 – PART D SAMPLING DESIGN

Selection of Elements • Population Total collection of elements on which we wish to make some inferences • Population Element Subject on which measurement is being taken • Sampling Selection of some elements so as to draw conclusions about the population • Census Count of all elements in a population.

Why Sample? • • • •

Lower cost Greater Accuracy of results Greater speed of data collection Availability of population elements.

What is a Good Sample? • Accuracy: absence of bias • Precision (Sampling error)

Types of Sampling Designs

• Probability • Nonprobability

Steps in Sampling Design • • • • • •

What is the relevant population? What are the parameters of interest? What is the sampling frame? What is the type of sample? What size sample is needed? How much will it cost?

Population Vs Sample ∀µ

__ X

• Proportion

∀π

p

• Variance

∀ σ2

s2

• Std. deviation

∀σ

s

• Size

• N

n

• Mean

Probability Sampling Designs • Simple random • Systematic • Stratified – Proportionate – Disproportionate

• Cluster (Area) (Single or Multi-stage) • Sequential / Double (2-phase) / Multiphase

ADVANTAGES AND LIMITATIONS ***

Simple Random Sampling (SRS) • Strengths – Easily understood – Results projectable

• Weaknesses – Difficult to construct sampling frame – Expensive – Lower precision – No assurance of representativeness.

Systematic Sampling

• Strengths

– Can increase representativeness – Easier to implement than SRS – Sampling frame not necessary

• Weaknesses – Can decrease representativeness.

Stratified Sampling • Strengths – Includes all important substations – Precision

• Weaknesses – Difficult to select relevant stratification variables – Not feasible to stratify on many variables – Expensive.

Cluster Sampling • Strengths – Easy to implement – Cost effective

• Weaknesses – Imprecise – Difficult to compute and interpret results.

Nonprobability Sampling Reasons to use • Procedure satisfactorily meets the sampling objectives • Lower Cost • Limited Time • Not as much human error as selecting a completely random sample • Total list population not available

Nonprobability Sampling • Convenience Sampling • Purposive Sampling (Judgement, Deliberate) • Quota Sampling • Snowball Sampling

ADVANTAGES AND LIMITATIONS ***

Convenience Sampling • Strengths – Least expensive – Least time consuming – Most convenient

• Weaknesses – Selection bias – Sample not representative – Not recommended for descriptive or causal research.

Judgemental Sampling • Strengths – Low cost – Not time consuming – Convenient

• Weaknesses – Does not allow generalisation – Subjective.

Quota Sampling • Strengths – Sample can be controlled for certain characteristics

• Weaknesses – Selection bias – No assurance of representativeness.

Snowball Sampling • Strengths – Can estimate rare characteristics

• Weaknesses – Time consuming.

Sample Size *** • When population can be quantified (Finite) • When population cannot be quantified (Infinite)

Infinite Population n = σ 2.z2 D2 D = +/- 5 (LOS = 5%) (also E) CL = 95% z = 1.96 σ = Standard Deviation n = estimated sample size

Finite Population n= D

σ 2. z2. N (N-1).D2 + σ 2. z2

= +/- 5 (LOS = 5%) (also E) Acceptable Error CL = 95% z = 1.96 Standard variance σ = Standard Deviation of population n = Estimated sample size N = Size of Population

1. 2. 3. 4. 5.

Sample Size Determination ***

Nature Universe (Dispersion Factor) Number of Classes Proposed Nature of Study Type of Sampling Standard of accuracy and acceptable confidence level 6. Availability of finance 7. Other considerations.

Nature of Universe • Homogeneous items – Small sample • Heterogeneous items – Large sample.

No. of Classes Proposed • Many groups and sub groups – Large sample.

Nature of Study • Items are intensively and continuously studied / Technical surveys – Small sample • General survey – Large sample.

Type of Sampling • Small Random sample better than large badly selected sample.

Standard of Accuracy & Acceptable Confidence Level • Higher Accuracy – Large sample. • If accuracy is doubled – sample size is increased fourfold.

Other Considerations • • • • •

Nature of units Size of population Size of questionnaire Availability of trained investigators Time available for completion.