false

Unlock Document

health sciences

NSG3

Hall- Fanshawe College

Winter

Description

Research final review
WEEK 10
Quantitative research:
Formal, objective, systematic process where numerical data is obtained
Describe, compare, predict a phenomenon
Descriptive- structured observations/survey, for phenomenon, situation or group
Exploratory- gain new insights, discover and increase knowledge about phenomenon
(with little information on the topic)
Casual- experimenting to assess cause and effect
Casual statements:
X precedes Y
X and Y are correlated
Everything else is controlled or eliminated. No extraneous variables impacting outcome
We never prove something. Gather evidence to support the claim
Qualitative Quantitative
Understanding Prediction
Interview/observations Survey/questionnaires
Discovering frameworks Existing frameworks (theoretical/conceptual)
Theory generating Theory testing
Textual Numerical
Quality of informant>sample size Sampling/sample core issues for reliability
Rigor: trustworthiness and authenticity Rigor: reliability and validity
Subjective Objective
Inductive Deductive
May be included in the research Excluded from the research
Research designs intro: Directs research process (who, what, when, where and how)
Located in method section
Purpose: answer questions, understand biases, direct analysis, direct interpretation
Develop study to bridge gap for what is unknown
Literature review is crucial-ethical approval and funding
Good designs-elements:
Objectively conceptualized- literature review reflect who what when where why
Accurate- answers research question accurately
Feasible- large samples to answer question within time and funding
Control- control potential bias to give valid results
Internally valid- results are believable
Externally valid- results are useful for other people/situations
Quantitative research designs:
Experimental:
o Randomized control trials
o Quasi-experimental
o Pre-experimental
Non-experimental
o Descriptive (correlational, univariate)
o Correlational (retrospective/prospective, natural, path analytical)
Models of analysis: parametric vs. non-parametric
Randomly assigned participants: experimental
No..
Control group/more than 1 measurement: quasi/pre-experimental
No.. non-experimental Experimental design:
Intervention that is controlled or delivered
There is an experimental and control group
There is random assignment to groups
* 1-3 true experimental
*1-2 quasi-experimental
*just 1 pre-experimental
Randomization:
Random assignment to groups (internal validity issue)
Equal extraneous variables in both groups
Random selection from population to sample (external validity issue)
Equal extraneous variables in sample that are true for population
Concept of variable:
Measurable characteristic that varies among subjects
Independent variable- interventions/presumed cause (salt intake)
Dependent variable- outcome/presumed effect (blood pressure)
Extraneous variables/covariate- alternative causes (exercise)
Concept of control:
Decrease error and the influence of unwanted extraneous variables
Increase probability of accurate findings and reflect true relationship between IV and DV
Control group: not exposed to intervention, eventually all groups exposed to the same
intervention Ways to control extraneous variables:
Use homogenous sample (same age)
Random assignment into groups (eliminate biases)
Minimize threats to internal validity
Use experimental design (manipulate independent variable)
Statistical manipulation
Concept of setting:
Natural setting- uncontrolled, real life situation (not recommended)
o Can influence results based on cause and effect
Partially controlled-manipulated/modified
o Quiet, light room to influence learning partially
Highly controlled- artificial environment for the research (recommended)
o Cause and effect strongly tested, results may vary in different setting
Research validity:
Design
o Internal
o External
Instrument
o Reliability
o Validity
Statistical
o Type 1 error (alpha)- false positive (mostly accurate, small amount by chance)
o Type 2 error (beta)- false negative (reject relationship based on the stats)
Internal validity:
Changes in the outcome (DV) due to changes in the exposure (IV)
Goal: rule out other expectation Threats:
o History- other concurrent events
o Maturation- developmental change
o Testing- pretest causes an effect
o Instrumentation- reliability of measure
o Mortality- subject drop out
o Selection bias- poor selection of subjects, no randomization)
External validity:
Are the findings generalizable to other populations/settings?
Goal: useful beyond the participants
Threats:
o Selection effects- sample does not represent population of interest
o Reactivity- natural reactions to being studied
o Measurement effects- act of being tested affects the outcome
o ** can exclude people based on the mean to not alter the results
Study designs to control threats:
Experimental (RCT)- controls most, gold standard
Quasi-experimental- controls some
Non-experimental- may control some
Descriptive- may control some
LEVEL SOURCE
1 Systematic review, meta-analysis, evidence
based guideline from RCT
2 RCT
3 Controlled quasi-experimental
4 Non-experimental
5 Systematic review of descriptive 6 Descriptive
7 Expert opinion
Design Threats t validity Strengths Limitations
Experimental RCT Selection bias Establish Difficult to
causality/casual implement
History direction
Generalizability
Testing
Control may be low
Instrumentation Not always
ethical
Measurement
effects
Quasi- History Establish Unclear cause
experimental causality/casual and effect
Testing
direction statement
Instrumentation Control Difficult to
implement
Measurement
effects
Generalizability
may be low
Not always
ethical
Classic No randomization step= quasi-experimental design
(non-equivalent control)
Non control group= quasi-experimental design (one
group, pre-test and post-test)
Non- Cross-sectional Selection bias Fast No causal
experimental association
Recall others Less expensive
Large # of
participants
Large # of
variables
Prospective Selection bias Time line VERY expensive History established Long term follow
up needed
Instrumentation Large # of
participants Large loss to
Testing follow up
Large # of
possible
variables
Retrospective Testing bias Fewer Difficult to find
participants adequate control
Selection bias group
Large # of
Selection effects variables
Causation Correlation
Time/temporality- cause must precede effect Statistical
Makes sense biologically/plausibility Needs corroborating research
Dose-response Association: General relationship between 2
random variables
Consistent association Correlation: linear relationship between the
random variables
Strength of association
Other quantitative studies:
Methodological- develop new tools/test existing ones in new populations
Meta-analysis- statistical: pool data from studies of same design
Secondary data analysis- descriptive: analysis or re-analysis of data for reason other
than why it was collected
Critiquing quantitative research:
Is the study design appropriate
Are the control issues discussed
Does the design reflect feasibility concerns
Does the design flow logically What are the threats to internal validity/how were they controlled
What are the threats to external validity/how were they controlled
WEEK 11
Terms:
Population- all possible subjects
Target population- all possible subjects who meet eligibility criteria
Sampling- sample selection
Sample- a subset of subjects
Element- one subject
Sampling:
Process of selecting study participants
Quantitative: to produce a representative group of participants so results from the
sample can be generalized to the population of interest
Sample:
people (children, seniors, patients), places (hospitals, units, cities), time (season, am/pm
shift)
heterogeneity- participants are diverse: wide age range, all types of cancer patients,
dissimilarities of the sample group inhibits the researcher’s ability to interpret findings to
make generalizations
homogeneous- participants are similar, all females, all same age, a group with limited
variation in attributes or characteristics
Threats to validity related to sampling:
selection bias- possible alternate explanation (internal)
attrition (mortality)- loss from groups (internal)
selection effects- sample does not represent population of interest (external)
Types of sampling strategies: probability (random)
o equal and independent probability of selection
o need complete frame
o not very common
strategies: simple random, systematic, stratified random, cluster
non-probability
o elements chosen non-randomly
o common
strategies: convenience (accidental), quota, purposive, network/snowball,
theoretical
Simple random sampling (SRS) -need a sampling frame (often incomplete or
out of date)
-each element has an equal & independent
probability of selection
-uses a random number generator
-is the bias for statistical inference
Systematic sampling -need sampling frame
-uses frame’s order to locate element (100
elements from frame 10, pick every 5 one)
-faster than SRS
-bias if frame arranged that coincides with
sampling
Stratified random sampling -need sampling frame
-used if SRS would provide too few elements
-divide sampling frame into strata than chose
random sample from each
-complicated and expensive but representative -proportional stratified: number chosen from
each strata proportional to sampling frame
Cluster sampling (multistage) -if sampling frame or each element is not
available
-uses stages of random sampling (select X
cities, select X hospitals in city, select X RNs in
hospital)
-convenient, efficient but higher sampling
error/bias possible
-more complicated analysis
Convenience sampling -use elements available at time and place of
study
-does not require sampling frame
-can be fast and efficient but may not provide
representative sample
Quota sample -researcher sets quota for number of elements
from specific strata
-ensures adequate elements from strata but
cannot ensure unbiased sample
Controlling Sampling error:
random selection
random assignment to groups or matching
estimate sample size using power analysis
overestimate sample size to account for drop out
Matching:
to reach pre-set sample size for groups
to ensure that important variables comparable in sample
often necessary for cluster sampling
harder to find elements/participants if matching-limits statistical analysis Sample size:
estimated a priori to determine number of elements required to demonstrate effect
too small=type 2 error or low power, non-significant result due to few observations
too large=unnecessary cost without added benefit of study results
critical piece of the protocol development
large sample size does not make up for non-representative sample
Sample size calculations:
not required for pilot studies
enlist the help of an epidemiologist or statistician
formula is based on:
o study design
o number of study variables
o sampling strategy
o heterogeneity of measures (literature review)
o precision required (treatment effect size)
o data analysis techniques (statistical tests planned)
o cost (time and money)
Power:
ability to detect a difference in effect between study groups or subjects (probability of
rejecting the null hypothesis)
not enough po

More
Less
Related notes for NSG3

Join OneClass

Access over 10 million pages of study

documents for 1.3 million courses.

Sign up

Join to view

Continue

Continue
OR

By registering, I agree to the
Terms
and
Privacy Policies

Already have an account?
Log in

Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.