Exam 2 Principles of Quantitative Research Flashcards
Design
Found in the methods section
Sampling plan
Found in the methods section
Data collection procedure
Found in the methods section
implementation strategies
Found in the methods section
data analysis plan
Found in the methods section
What is a design?
Specific features need to be addressed, need to be structured. Step by step
approach. Different kinds of sampling, different kinds of data collection.
Blueprint
says what to collect, how to collect it, and what you’re going to do with it
What is a good design
Appropriate to the purpose
-subjects, setting, protocol and comparisons
-most importantly, must be able to answer the question
Appropriate comparisons
-Also needs to be feasible given realistic constraints
-availability of data…need to be able to find the setting and subjects
Design must also
be effective in reducing threats to design validity
guard against negative influences that may negatively impact data
Major categories of design
Experimental
Most “scientific” gives best information, best evidence
Contains hypothesis. Manipulating variables. cause and effect situation
Make comparisons between real people/real groups that have a real situation occuring
Major categories of design
quasi-experimental
something is preventing it from being a true experiment. ethical considerations. So simulate the “experiement
Major categories of design
Nonexperimental
Not meant to test a hypothesis. meant to describe by collecting numbers, exploring phenomena by collecting. situations in natural habitat. can’t provide hard data…assumptions made
Causality
Basis of cause and effect.
There must be a strong correlation or relationship between I.V. and D.V.
Cause must precede effect
Must be necessary and sufficient (must always be present and requires no other factors)
Must be no alternative explanations
Very hard to prove
Multi-causality
few events of interest to nursing have a single cause
These other multiple, interrelated causes need to be controlled.
Probability
Research on living beings tends ot be multicausal
- Impossible to determine true cause and effect
- probability is best guess given the current information
- addresses relative rather than absolute causality
- a probable explanation is more in keeping with multicausality
control
-to control means to have the power to direct, regulate, manipulate, or statistically adjust factors to achieve a desired outcome on the D.V.
Ways to achieve control
- Control the environment
- Control equivalence of subjects and groups
- Use control or comparison groups
- Control the treatment
- Control the measurement
- control extraneous variables
Manipulation
Important principle of control (alter the I.V. by increasing, decreasing, withholding, adding, etc the I.V.)
- True experiments must be able to control the intervention or treatment being tested.
- Intervention group vs. control group
Bias
is anything that causes a deviation from the truth or that distorts findings. (The results are slanted in some fashion)
-Can be caused by extraneous variables (control variables/confounding variable)
Randomization
A method to diminish bias (gives a more realistic picture of the population)
- Distributes confounding factors across the whole group as they naturally occur by chance rather than clumping those factors in any one part of the group because of errors in the design (site selection/subject selection/group assignment)
- differences between groups causes a need for statistical adjustment
Group comparisons
Between groups
Compare two or more whole groups (ex: men and women)
Group comparisons
Within groups
Compare members of one group (ex. just look at males or females. May also look at ONE person pre and post intervention)
Validity of the study design
Truth or accuracy of the results
Meaning, are the results:
Logical, reasonable, justifiable
-must refer back to theoretical basis and propositions
Types of validity
Internal, external, statistical conclusion, and construct
Internal Validity
Does the manipulation of the I.V. REALLY make a significant difference in the D.V.
Or, is a rival hypothesis correct?
Threats to internal validity
Selection bias history maturation testing instrumentation mortality statistical conclusion validity
Threats to internal validity
Selection bias
Characteristics of the subjects are not evenly distributed across the whole group (subjects are too different…some subjects have certain characteristics that are not normal in the group
- this makes it hard to make generalizations
- to prevent use randomization or homogeneous groups if non-random, matching subjects, or statistical control
Threats to internal validity
History
- Something happens during the course of the study that impacts the results
- Can use a control or comparison group to help identify this
Threats to internal validity
Maturation
- subjects naturally change over a period of time
- changes are developmental and not due to the study I.V.
- use a control group to detect this
Threats to internal validity
Testing
- Comes into question with pre-test/post-test designs
- People may remember the questions from the re to the post test
- Pretest may cue subjects into what to focus on
- Control by extending the time between the pre-test or post-test
Threats to internal validity
Instrumentation
Tools/instruments/questionnaires must be accurate
tests based on observer input need to have observations recorded in the same manner (interrater reliability)
Threats to internal validity
Mortality
Loss of subjects (attrition)
- Who were they?
- Were the different than the ones you retained?
- Why did they leave/die?
- Should be reported in the results.
How to control mortality?
Have shorter studies or taking good notes on who left and why
Statistical Conclusion Validity
Based on our statistics
- Concerned with whether the conclusions about relationships or differences drawn from statistical analysis are an accurate reflection of the real world.
- Are we confident that we are correct in our findings?
- Look at reported confidence intervals
- Concerned with power, test assumptions, error, reliability of measures, treatment implementation and random differences in subjects.
Power
The ability of a design to detect true relationships among variables
- Needs a sufficiently large sample for the statistical tests being run
- with randomization, maximum differences in groups on the I.V. (treatment) will result in maximum effects on the D. V.
Measurement errors
Type I
I.V. and D.V. in our study are related, but not related in the real world
Measurement Errors
Type II
I.V. and D.V. in our study are not related, but ARE related in the real world
Construct validity
is there a good fit between the conceptual definitions and the operational definitions of the variables?
- Check the theory for the conceptual definitions
- Check the description of the concept given by the developers of the measurement tool
- does the measurement tool actually measure the concept as proposed in the theory
- do multiple tools used to measure the concept all get the same results?
External validity of the design
Can you validly apply these findings to people who were not in the study?
-can you replicate the results?
Threats to external validity
Population Validity
Interaction of treatment and subjects
subjects can sometimes be influenced by participating in the study
-changes noted in D.V. are due to artificial subject reactions rather than the I.V. (Hawthorne Effect)
-sample may not be representative of the larger population
-researcher may “stack the deck” as subjects are selected
-sample selection may pick from too narrow a group
Threats to external validity
Ecological validity
Interaction of treatment and setting
setting issues
-may not be a “normal” setting
-data collected at different settings may be very different (distractions may influence attention/cultural differences)
-time and day data is collected (fatigued subjects)
Threats to external validity of the design
pretest sensitization
-interaction of treatment and history (change in D.V. is due to confounding by an event in the past rather than the I.V…..remembering pre-test question)
Time related design categories
Retrospective
Cross-sectional
longitudinal
Time related design categories
Retrospective
- research is conducted after the fact (ex post facto)
- I.V. occurred in the past
- D.V. is being measured now
- Tries to find cause of current situation
- May look at subjects where D.V. is present and compared to subjects where D.V. is not present
Time related design categories
Cross-sectional
-Gather all data about I.V. and D.V. at one time or in a short time span (several hours)
-May pick subjects with a wide range of the concept of interest right now (ex. recruit 100 people with diabetes…then divide into newly dx, dx 5 yrs, dx 10 yrs, dx 15 yrs. Pretned that we are looking at ONE group of people over time)
May deliberately pick several groups to compare
Time related design categories
Longitudinal
Prospective in nature (looks at chronological change)
I.V. is introduced then D.V. is measured at a later date
May follow same subjects into the future (high school GPA and college GPA)
May follow different groups over time (hospital fall rates)
Ethics
Focus on control, manipulation and prevention of bias
- Selection of subjects
- Assignment of subjects
- Selection of treatment
- Assignment of treatment
Time and money
Retrospective, current, prospective
