Why We Need Sample

 Why We Need Sample

13/2

Population is too large
Cost, time, human resource cinsideration
Destructive case
Carefully chosen sample can be used to represent population

Sampling frame is list of sampling frame

Sample is collection of sampling units drawn from sampling frame

Parameter: numerical statistic of population

Statistic: numerical characteristic of sample

Basic concept in sampling
We draw sampling from population parameter.
Sample statistics
Sample inference

Beware with sample frame error and sample error

Sampling error is statistical error that occur when selecting sample that not represent entire population

Result found in sample thus not represent the results that would be obtained from entire population

Sampling error can be reduced by randomizing sample

Deviation

Non sampling error: error occurs during data collection. Causing data differ from true value
Include non response error, coverage error, interview error, processing error
When non sampling error occur, rate of bias in study goes up

Sampling frame not match perfectly with target population, leading error of coverage

Missing observation

Duplication

Erroneous inclusion

Non response error is the most serious problem

Error:
Respondent
Instrument
Method of data collection

Train the enumerator/interviewer, not the responden

Make sure they are qualified and capable to deliver our questionnaire and message

To re-check:
Use same responden with different enumerator

If we want to take model, check the reliability

In usual situation, the model is normal

Increase samples to make data become more reliable

2 situation:
Parametic statistic
Non parametic statistic

Probablity vs non probability sample

Probability = randomly choose
Get equal opportunity to be selected as sample
Method include simple random sampling

Non probability samples
Members are selected from population in some random manner

Method include convenience sampling, judgment, quota, snowball sampling

Key
Meaning: equal opp -wherein
Alternatively known as: random-non
Basis of selection: random-arbitraly
Opportunity of selection: fix known - not
Research: conclusive-explatory
Result: unbiased- biased
Method: objective-subjective
Inference:stat-anytc
Hypothesis: tested-generated

Simple random sampling: purest
Each member of population equal
Each possible sample equal

Use software ie. Excel, minitab, random generator etc

Systematic sampling
Nth name selection technique

After required sample size has been calculated, every Nth record is selected from a list population members

As long as the list not contain any hidden order, sampling method is as good as random sampling method

Advantage of random sampling

Determining numbers of interval
As long as interval not larger than population, it's fine

Make sure the orders are known by researchers

Make sample systematically based on the interval

When is the appropriate to use random sampling and non-random sampling for quantitative survey and qualitative survey?
how to decide the number of sample for population that we don't have information about the number of population? For example, like case of indigenous people

Cluster sampling: Sample by cluster

Effective in condition
Good sampling frame is not available or costly

Ex:
City block
Street
Housing unit
Hospital
Automobile

Multi stage sampling
Use smaller sampling unit at each stage

Combination stratified/cluster sampling and simple random sampling is usually used

Complex form of cluster sampling

Non probability sampling

Convenience sampling
Used during preliminary research
Used in exploratory research
Useful for pilot studies

Sample is selected coz they are convenient

Judgment sampling
Nonprobality method
Sample based upon judgement
Extension of convenice sampling

Researcher must be confident that the chosen sample is truly representative of entire population

Quota sampling ~ stratified sampling
Identify stratum and their proportion as they are represented in population

Convenience/judging sampling is used to select
Ex: male, above 50

Snowball sampling used when desires sample characteristic is rare

It may difficult or cost

Rely on referral like MLM

Lower search cost but can bias

Sample size?

Slovin formula
Cochran formula

the cost of obtaining observation increase as the distance separating the elements increase. So, is the parameter of obtaining observation cost only based on distance?

is convenience sampling possible to be called targeted sampling as well?