Exam 2 Quantitative Sampling Flashcards
Population
group meeting a specific description
Target population
everyone who meets the criteria (very specific criteria for study)
accessible population
those who meet the criteria and who can be accessed by the researcher (who researcher has access to)
Sample
selected portion of group you are studying
-small, specifically selected to be in study
Elements
individual or basic units to be studied
subjects/participants
human elements to be studied
Sampling plan
how you will select the sample (number, eligibility, process) written description
What is representativeness?
the sample looks enough like the population to be able to GENERALIZE the sample findings to others in the population
How must the sample similar to the population
on important characteristics
Inclusion criteria
characteristics that the subject MUST possess to be in the study.
Based on notable characteristics of the target population
Exclusion criteria
characteristics that the subject MUST NOT possess
reduces questions about impact of extraneous variables on the results
sampling bias
over or under selection of subjects on any of the significant variables
sampling error
errors made in the conclusions because of sampling bias
Clinically relevant findings
Overall picture must be one closely resembling
Probability sampling
Every element(person) in the accessible population has an equal chance of being selected for inclusion in the study
3 conditions:
accessible population must be identifiable (must be able to make a list)
All elements in the accessible population must be listed (sampling frame)
Elements are chosen by random selection
Selection method is determined by:
research purpose, question, and design
Types:
simple, stratified, cluster, and systematic random sampling
Randomization
outcome of random sampling methods
What does randomization reduce threat of?
selection bias
Less selection bias
ensures a greater likeness between the sample and the target population on the focal characteristics
Common methods:
- Draw names out of a hat
- Use a table of random numbers
- Flip a coin-heads in, tails out
- Throw dice-anyone getting an even number is in, odd number is out
Simple random sampling
everyone in the accessible population has an equal chance for inclusion
process:
-get a list of the accessible population
-assign everyone on the list a number
-randomly select from the list
…whatever it takes to remove brain’s ability to pick
Difficult?
if it’s a large group, or when one cannot make a sampling frame
-to use when one does not know who will be admitted to the unit/hospital/etc meeting the focal characteristics
stratified random sampling
select the relevant strata
(must be mutually exclusive)
-divide the population into subgroups or strata based on some important feature.
-decide on how many from each strata
-use random sampling to select that number from each strata
Difficult?
when accessible population for one strata makes it small or difficult to find
Cluster or multistage sampling
clusters narrow down large unwieldy population (Still RANDOM)
- Identify important characteristics
- look for logical subsets (clusters) of the large population
- randomly select your clusters and sub-clusters
- randomly select subjects from you clusters
systematic sampling
create a list of the accessible population
-divide the number of people available by the number of subjects needed to get ‘k’
example: 500 people/50 needed=10
-draws every ‘kth’ element
example: the above answer of 10
means every 10th person on the list
-randomly pick a starting point on your list\
-select every kth person (10th person) until you have your sample.
nonprobability sampling
less likely to give you a representative sample
-increased seelction bias
when use nonprobability?
when sampling frame cannot be determined
types:
convenience, quota, purposive or judgement, snowball, theoretical, expert
Convenience sampling
Use any subjects
- whatever is available to you
- volunteers
ex: mall surveyors
Convenience sampling: control sampling bias
collecting data on subject characteristics and then making comparisons of those characteristics
Quota sampling
- similar to stratified random sampling WITHOUT the random selection steps
- decide on variables of interest
- decide on the strata to select from
- conveniently select from each strata (instead of randomly select)
Purposive sampling
- base selection on specific criteria
- inclusion relies on researcher’s judgment
When is it used?
More often used in qualitative than quantitative designs but may be used in quantitative descriptive studies
Snowball sampling
A type of purposive sampling
How are subjects obtained?
By referral
How is this useful?
Helps to obtain key people or those who might be difficult to find under normal circumstances
Theoretical sampling
Used mostly when trying to develop an explanation of a phenomena
Two step process
- Collect data
- develop initial explanation (theory)
- purposefully identify additional subjects to determine if the explanation (theory) continues to hold up
Expert sampling
Use when looking for consensus on an issue
When to use?
Creating research tools
It helps measure content validity
Sample size
Samples that are homogeneous (similar) on the desired characteristic do not need to be as large as samples that are heterogeneous (dissimilar) on the desired characteristic.
How does reliability/validity effect sample size
The more reliable and valid the tools, the smaller the sample size needed
General rules of thumb for sample size:
Minimum of 30 for one group study
- if doing group comparisons, need at least 30 in each group
- if more than one I.V., need at least 10 for each I.V.
Also need to consider for sample size:
Refusal rate since not everyone will want to participate
- consider your likely “drop out” or attrition rate
- consider “power” of your study
Statistical power
looks at your assurance that you are correct in rejecting the null hypothesis
- need to consider your alpha (significance lvl), your effect size, and desired power level
- if you set your alpha at .05, you are looking for a 95% chance of correctly rejecting the null hypothesis
- must decide if you think the I.V. will have a large, medium, or small impact or effect on the D.V.
- generally wish a power level of at least .8
- use your alpha, effect size, and desired power to calculate the number of subjects you need for each individual statistical procedure used in the study
Recruitment and retention
IRB approval required before approaching any potential subject
How can you recruit?
By using personal contact on the unit, contact manager or staff member on the unit, flyers, ads on radio or TV…etc…
Requirements for recruitment
All printed documents must be at maximum of 5th grade level in appropriate language
- incentives should not be bribes
- determine who meets likely the criteria
- be professional, personable, and patient
- explain exactly how they will participate such as in person, survey monkey, mail in responses, etc…
- provide for informed consent
- have informed consent form signed if needed
Do you always need a SIGNED informed consent form?
NO
Informed consent must include:
- title of study
- invitation to participate
- basis of subject selection
- purpose of study
- explanation of procedures
- benefits and risks
- alternatives to participation
- financial obligations
- confidentiality assurances
- HIPAA disclosure of how their data will be protected
- subject withdrawal procedures
- offer to answer questions with contact information provided
- consent statement
- identification of researchers
- signature and date, if required
Who are the Vulnerable Populations?
- Pregnant women
- Fetuses
- Children
- Elderly
- Mentally challenged
- Prisoners
- Other marginalized individuals based on socioeconomics, ethnicity, gender, education, etc.
