SAMPLING TECHNIQUES Oyindamola B. YUSUF Biostatistician KAIMRC-WR
LECTURE OUTLINE• Introduction• Determining number of subjects• Total population Surveys• What is sampling?• Reasons for taking a sample• Principle of sampling• Types of sampling
DETERMINING NUMBER OF SUBJECTS• Statistical considerations• Practical Considerations
PRACTICAL CONSIDERATIONS• AVAILABILITY OF SUBJECTS• RESOURCES- TIME, MONEY, PERSONELL
STATISTICAL CONSIDERATIONS• Purpose of study• Primary outcome measure• How small a difference is to be detected• Type 1 error: To find treatments significantly different if treatments don’t really differ• Drop out rate if study is prospective• Power to detect an actual difference
TOTAL POPULATION SURVEY• Every individual in the defined population is included and studied.• It has Advantages and Disadvantages
ADVANTAGES OF STUDYING TOTAL POPULATION• (a) The estimate is accurate and without error since no unit is left out• (b) There is no need to worry about selection procedure• (c) And there are no feelings of discrimination created in the population.
DISADVANTAGES OF TOTAL POPULATION STUDY• (a) It is expensive• (b) It takes time to complete• (c) Demands a lot of personnel• (d) It may not be feasible• (e) It may be less accurate.
WHAT IS A SAMPLE ?• Part of a population selected for study• May be able to infer the characteristics of the population from those of the sample.
ADVANTAGES OF TAKING A SAMPLE• Advantages• (a) Less expensive• (b) Quick results guaranteed• (c) Demands on personnel is less• (d) Possibility of obtaining more accurate data because of the smaller number of units involved
DISADVANTAGES OF TAKING A SAMPLE• Estimate obtained from the sample is likely to be different from that would have been obtained if the total population have been studied. – - this discrepancy is called sampling error and it is always present.• It is sometimes difficult to select a good sample i.e. a representative sample.
GENERAL CAUSES OF BIAS• a. Lack of proper knowledge of the population from which the sample is selected.• b. Inadequacy of sampling frame.• c. Personal prejudice - i.e. when personal feelings is allowed to influence sample selection-observer error.
HOW TO AVOID BIAS• TAKE A PROBABILITY SAMPLE• THIS IS KNOWN AS A RANDOM SAMPLE• SAMPLE HAS A KNOWN CHANCE OF BEIGN SELECTED
DEFINITIONS OF TERMS NEEDED TO TAKE A PROBABILITY SAMPLE• (i) Sample Size• (ii) Sampling Fraction• (iii) Sampling Frame• (iv) Sampling Unit• (v) Unit of Enquiry• (vi) Sampling Error• (vii) Good or Representative Sample
EXAMPLES OF PROBABILITY SAMPLES • SIMPLE RANDOM SAMPLE • SYSTEMATIC SAMPLE • STRATIFIED RANDOM SAMPLE • CLUSTER RANDOM SAMPLE • MULTI-STAGE RANDOM SAMPLE
SIMPLE RANDOM SAMPLE• Simple random sample: A sampling procedure in which each unit in the population has the same (equal) chance of being selected. However the population must be finite and a sampling frame must exist.• Each unit must have an assigned number in the sampling frame. Without a proper sampling frame, it is impossible to take a simple random sample.
SELECTION PROCEDURE OF SIMPLE RANDOM SAMPLE• 1. Lottery method• 2. Use of table of Random numbers.• 3. Use of computer facilities.• Lottery• 1. Construct a frame of all the sampling units.• 2. Use ballots to select the required number of units.
SYSTEMATIC RANDOM SAMPLE• Unit selected in any one sample occupied related position to each other in the sampling frame• Determine sampling fraction and sampling interval-k• The first unit to be selected is selected at random between 1 and k.• Thereafter every kth unit is selected.
EXAMPLE ON SYSTEMATIC SAMPLE• Suppose a sample of 50 patients is required from the register of 1,000 patients available in the records section of a teaching hospital. The sample fraction here will be 50/1000 = 1/20 , thus k = 20.• The first member in the register is selected randomly between 1 and 20.• The first and every 20th member is subsequently selected as sample members.
STRATIFIED RANDOM SAMPLE• Population is divided into homogenous strata according to some relevant characteristics of the population• A random sample is selected from each stratum• The sample size may be sub-divided in proportion to the population size in each stratum. This is called a proportional allocation.• For example to select 200 units from a population of 6000 units of which 2000 units are females and 4000 units males. The number chosen in each sex stratum will be 68 and 132 respectively if there is a proportionate allocation of the sample numbers in the strata.
MULTISTAGE RANDOM SAMPLE• Multistage Sample• Sampling in stages• Final sample obtained after more than one stage• Ex. Selection sample of students from the university• 1st Stage Selections: Select 50 depts at random out of the existing 160 (for example).• 2ND Stage: from each selected depts, list all the students then select• students in each of these.
CLUSTER SAMPLE• The sampling unit is a cluster of units• Units could be households, streets, or villages.• The approach is useful in rural areas where there are no sampling frames.• Multi stage sample and cluster sample are the most popular method in a rural area.• In cluster sample, selection takes place only once.
EXAMPLE OF CLUSTER SAMPLE• Study of attitudes of medical students to HIV/AIDS pandemic.• Define each department as a cluster of students• May select a specified number of departments at random out of the total number of departments in the University of Ibadan• Study all students in the departments selected.