Module 8 - Exploratory Research and Correlation Flashcards
Uses of Exploratory Research
Estimate risk factors; (factors that may increase or decrease likelihood of developing a condition)
Explore relationships between two or more variables; (correlational research)
What is one of the most common uses of an observational design?
To draw causal inferences about the effects of a hypothetical RISK FACTOR
Causality
The goal of many observational studies is to IMPLICATE factors that contribute to health outcomes
If there is NO control/manipulation or randomization in an observational study, _________.
causality is more difficult to determine
Which is the only ethical and practical methods of investigating the causal impact of a potential risk factor?
Sometimes observational designs is the only one
Observational Study Approaches:
Longitudinal studies
-Prospective (looking forward in time)
-Retrospective (looking backward)
Cross-sectional studies
Longitudinal studies and Challenges in design
subjects are followed through time; can be performed either prospectively or retrospectively
Prospective study
- Long time period to consistently collect data
- Exposure status determined at start, with follow-up to see if outcome develops
- Confounding can occur over time from other events and conditions.
Retrospective study
- Involve the examination of previous data (e.g., medical records/notes, databases, existing survey etc.)
- Difficulty in defining variables
- Records may not be complete.
Cross-sectional studies
- Considered a “snapshot” of a population
- Data for exposure and outcome taken concurrently
- Can describe health status of a population at a point
in time
Cross-sectional studies and Challenges with design
Due to a lack of the time sequence, it may not be
possible to know whether the presumed cause
(exposure) truly proceeded the outcome.
Cohort studies (follow-up study)
Longitudinal investigation
* A group of individuals are followed over time
✓Subjects classified by exposure status at the start
✓All have potential for outcome condition
* Study association between exposure and the
outcome of interest
✓Outcome determined over time
* Can be either prospective or retrospective
Challenges for Cohort studies
- Misclassification of exposure: over-report or under-report of exposure
- Attrition and bias
✓ due to longitudinal nature, prospective cohort studies are prone to attrition, and
✓ Result in bias if the loss is related to exposure status, outcome status or both - Outcome may not occur in sufficient numbers
✓ Not appropriate for studying rare conditions
Case-Control Studies
A method of observational investigation in which groups of individuals are purposely selected on the basis of whether or not they have the health condition under study
Cases
those with the target condition
Controls
those who do not have the target condition
Caso-Control Useful for?
Useful for studying rare or conditions with long latency
Exposure status compared between cases and controls
Challenges for Case-Control Studies
Selection bias
Observation vias
Recall bias
Confounding
Selection bias
is a special concern, because subjects are purposefully selected based on whether a target disorder is present or not (especially when volunteering to participate)
Observation bias
Foreknowledge of outcomes may render bias by assessor or biased reporting by participants
Recall bias
Those with a condition may have better recall of exposures
Confounding
confounding effect of extraneous variables that are related to the exposure of interest (something that is known already/ using someone who is more likely)
Correlational Research
Purpose: to examine relationships between /
among two or more variables
can be either retro or pros
Correlational Research design
Usually involves one group of subjects with
measurements taken of different variables in order
to establish a mathematical relationship among the
variables
* To predict scores on one variable based upon
scores on another variable (predictive correlational
studies)
15
Correlation Coefficient
- Indicates the strength of the relationship
✓ Values between –1.0 and +1.0, where 0 is no relationship - Sign implies direction of the relationship
- H0: population correlation = zero
Assumptions for correlation
Scores represent the underlying population
* Scores are normally distributed
* Each subject has a score for both X and Y
✓ (e.g., X = GPA & Y = clinical performance score)
* X and Y are independent measures
* X and Y values are observed, NOT controlled/manipulated
* Relationship between X and Y is linear, not curvilinear
Strength of correlations
Interpretation affected by sample size, measurement
error, types of variables being studied, and their
application.
* Authors should provide rationale for interpretation.
Pearson Product Moment correlation
Parametric test that is appropriate to use when:
* X and Y continuous variables
➢ Interval or ratio scales
➢ Normal distribution
* Sample statistic: r statistic
* Population parameter, ρ (rho)
Null H0: = 0 Alt H1: ≠ 0
* Clinical vs. statistical significance
Spearman Rank correlation coefficient
- Nonparametric analog of Pearson’s r
✓ Based on ranked data
✓ Ordinal data or non-normal distribution - Symbol, rs
Null H0: rs = 0 Alternative H1: rs > 0
Correlation versus Comparison
- Interpretation based on covariance
✓ The change in X is proportional to the change in Y. - Must be distinguished from the determination of
differences between two means/distributions (in which
t-test or ANOVA applicable)
Correlation and Causation
- Presence of association does NOT imply a causal
relationship
✓ Strong relationship between X and Y may be a function of a third variable
Outliers
- Data points that lie outside the cluster of scores
- Examine using a scatter plot
- May be an extreme score due to a small sample size
or may result from measurement error or from other
extraneous factor - Need to determine if outliers should be retained or
discarded
Range of test values
- Generalization of correlation should be limited to the
range of values used to obtain the correlation - Should not extrapolate to scores outside this range
