4130 Chapter 12: Sampling

*Is a process of selecting representative units of a population for study in a research investigation.

*A whole population cannot be studied thus a smaller sample must be obtained.

*To ensure that scientific findings are as accurate or true as possible the researcher must put in controls to match the sample to the larger population.

*A population is a well-defined set that has  certain specified characteristics from which data can be gathered or analyzed

• (people, objects, animals, all the children admitted to hospital
• in Alberta in 2008).

Sampling: subset of the population that contains elements

• Elements: basic unit about which information is collected
• -people
• -places
• -objects 

When sample is planned...the researcher must...

Specify inclusion (eligibility)

Specify exclusion (delimination) 

=leads to sample selection

Examples of descriptors are: 

gender, age, marital status, socioeconomic status, religion, ethnicity, education, health status, diagnosis, co-morbidities or BSN students. 

Examples: population descriptors

Think about the concept of inclusion or eligibility criteria applied to a research study where the subjects are clients. For example, participants in a study investigating the pain experience of children with leukemia during the first year after diagnosis had to meet the following inclusion (eligibility) criteria

*Age: 4 to 17 years.

*Diagnosis: acute lymphocytic leukemia within one month of diagnosis.

*Health status: no other existing chronic illness associated with pain, fulminating disease, or known cognitive disability.

*Status: ability to cope with the burden of research tasks as determined by the primary nurse.

*Language: English- or Spanish-speaking.

Homogeneity: same characteristics in a sample

Heterogeneity: dissimilar characteristics in a sample. This decreases researcher's ability to increase generizability (↓theresearcher's ability to generalize research findings to the larger population [external validity]) The process is a control to ↓sample bias

Target population: 

is the entire set of cases about which the researcher would like to make generalizations. An example: all the BSN students enrolled in a BSN program in Canada

Accessible population:

meets the researcher's target population  criteria by is available to the researcher. In our example it may be all BSN students enrolled in a BSN program in BC

representative sample is one whose key characteristics closely approximate those of the population

Nonprobability:

nonrandom. All sample characteristics DO NOT have a chance to be represented

Probability:

random selction of the sample units. More likely to be representative of the population. Rigorous sampling strategy

*Randomization or random assignment is the distribution of subjects to either an experimental or a control group on a random basis. This refers to the groups within the sample.

*Random selection is when each element of the population has an equal and independent chance of being included in the sample. Random selection refers to the larger population not the sample.

*These are not the same concepts. An example is that a study may use a convenience sampling method (non probability type) and randomize the subjects into a treatment or control group.

*A randomized clinical trial (RCT) is a full experimental test of a new treatment, involving random assignment to treatment groups & is of a large & diverse sample (aim is for random selection sampling).

*Note in a research report how the researchers sampled the larger population to determine if it was representative.

Convenience: quant, very easy drawing representative sample, greater than any other sampling strategy for risk of bias, representativeness is questionable

Choosing the most readily accessible persons or objects as study subjects. Accessibility = convenience for the researcher.

High level of research bias thus ↓internal & external validity

(Iint valid: if the IV caused the change in the DV [hx, maturation, testing, instrumentation, mortality, selection bias. Ext valid: which the findings can be generalized [subject selection, reactive effects or study conditions, measurement/testing effects]

examples:

*the first 24 patients admitted to ICU at hospital x with pneumonia.

*the patient's on a oncology unit. 


Quota (def: a nonprobability sampling strategy that identifies strata of population and proportionately represents strata in example): quant, relatively easy

• *A strata of the population is known & thus a strata
• is chosen for a sample. NOT randomly selected.

The problem is that there could be over or under representation of characteristics. Based on researcher judgement thus bias. 

*This provides the highest level of sampling strategy.

• *Simple random
• *Stratified random
• *Multistage (cluster)
• *Systematic

A. Simple Random: laborious, low risk of bias, representative of sample maximized

 

Advantages and Disadvantages:

*No research bias.

*Generalizability to the larger population is maximized.

*Differences between the sample & population are purely chance.

*The larger the sample size the greater the representativeness or study external validity.

*Disadvantage is that it is timely, costly & pose issues in obtaining accurate records/lists of every element of the population.

*Reader be warned: Every type of sampling has some drawbacks. Look for how the researcher addresses them in their report. Ask why the sampling type was chosen & if it is randomized is it random selection or random assignment. 

•  
• B. Stratified Random: time-consuming, low risk of bias, enhanced representative of sample 

*The population is divided into homogenous strata or subgroups.

*The proportions of the sample represent the proportions noted of the concept being studied as it presents in the larger population.

*The aim is to get a representative sample.

*The concept is the same as nonprobability quota sampling but random selection of the larger population is used.

*It is difficult to find a population list containing complete information, it is time consuming, enrolling a proportion of the population is challenging & costly. 

C. Multistage (cluster): less time-consuming than simple/stratified sampling, subject to more sampling errors than simple or stratified sampling, less representative than simple or stratified sampling

*This involves successive random sampling of clusters that progress from large to small until the sample criteria is met. 

Example:

*Clinical nurse gerontological specialists are desired as a sample.

*1st sample: random sample of hospitals where CNS work (a list obtained from the CRNBC).

*2nd sample: CNS in each hospital from the 1st sample (a list from the HR department at each hospital). Random selection of 2 CNS from each hospital who meet the eligibility criteria set by the researcher. 

D. Systematic Sampling: more convenient & efficient than the other 3, bias in the form of nonrandomness can be inadvertently introduced, less representative if bias occurs as reult of coincidental nonrandomness

*A strategy where a subject is chosen at fixed intervals such as every 10th person (or 5th, 7th, 10th)

*This is to be determined prior to the study.

*The aim is to get the largest sample possible because it strengthens generalizability. 

*A sample size in quantitative research can be determined by a statistical procedure known as a power analysis. Researchers refer to this as power. The goal is to get a representative sample. 

• *To determine this the researcher must first determine
• how much effect the treatment will have on the outcome of a study. For example a study to see the effect of pre-operative teaching on patient post operative anxiety.

*The difference between anxiety pre ope & anxiety post ope is called the effect size.

• If a small difference (effect size) is expected the
• sample must be large.

When a large difference (effect size) is expected the sample can be smaller

_________________________Review______________

*This is to be determined prior to the study.

*The aim is to get the largest sample possible because it strengthens generalizability.

*A sample size in quantitative research can be determined by a statistical procedure known as a power analysis. Researchers refer to this as power. The goal is to get a representative sample.

• *To determine this the researcher must first determine
• how much effect the treatment will have on the outcome of a study. For example a study to see the effect of pre-operative teaching on patient post operative anxiety.

*The difference between anxiety pre ope & anxiety post ope is called the effect size.

• *If a small difference (effect size) is expected the
• sample must be large.

• *When a large difference (effect size) is expected the
• sample can be smaller.

type of design used

type of smapling procedure used 

type of formula used for estimating optimum sample size (power analysis)

degree of precision required

heterogeneity of attributes under investigation

relative frequency that the phenomenon of interest occurs in the population (ie. common vs rare health problem)

projected cost of using a particular sampling strategy