Exam 3 Flashcards
Sampling, study designs, correlation vs regression
What is the difference between a study completed using a sample and a census?
A study using a sample is a representational portion of the total population while a census is meant to be a study of the ENTIRE population
What is sampling error?
When the sample does not properly represent the population.
What is the difference between biased (non-probability) and representative samples (probability)?
A biased sample is more typically used in research due to the difficulty of creating a truly representative sample (availability of college students vs the entire population).
How do researchers create biased samples?
Self-selected sample (volunteers), snowball sampling (word of mouth), purposive sampling (cluster sampling, targeting a specific group, posting study fliers within a gay bar), quota sampling (stratified sampling, requiring a specific amount of specific participants)
What is a multistage sample?
Sampling participants in multiple stages, choosing a random sample of clusters and then randomly sampling from the chosen clusters
Why are representative samples better than biased samples?
Representative samples have higher external validity due to being perfectly representative of the population.
What is cluster sampling?
Biased, choosing clusters participants within a population (pre-k teachers vs high school teachers)
What is a stratified random sample?
Choosing a specific characteristic to sort the population (male vs female) and making sure that this characteristic is properly represented in the sample of the population
What is the benefit of oversampling?
When sampling a minority population, it may be importance to focus on gathering minority participants instead of the majority (When studying non-binary individuals, it is important to ensure that the sample is representative of the non-binary population through oversampling)
Define external validity.
When the concept being studied can be broadly applied to the population (hours slept and how this impacts all human tasks).
Define internal validity.
When the concepts within a study interact and influence one another (hours slept vs score on math test).
Define statistical validity.
When a study is confirmed through statistics.
Define construct validity.
How accurately the construct is defined and measure.
What shows the strength of the correlation?
How close ‘r’ is to 1.0
Does the sample size impact external validity?
No, the method of sampling impacts the external validity much more than the overall size
What is the point of ‘r’?
The r value shows the strength of the statistical relationship between variables (SES vs Level of Education)
Does the type of statistical analysis impact the claim of the study?
No, the type of test does not determine the type of claim a study can make (causal claims).
What types of studies typically use a t-test for statistical analysis?
Manipulation/experiments
What types of studies typically use a correlational test for statistical analysis?
Measured variables/association
Why don’t the typical rules for understanding a correlational analysis apply to a curvilinear result?
The results are significant on both sides of the data (serial position effect)
What are spurious correlations?
A spurious correlation is a possible outlier influence/reason for the significance of the correlation that does not include the two variables
Define moderation.
When the relationship between two variables changes depending on the level of the third variable (shows different associations between original variables, maternal employment influences their child’s vocabulary ability, but the mother’s marital status also has a big influence)
Are spurious correlations and moderator variables related?
Yes, they can be one in the same, (kidnapping rates and number of vacations taken within one year, proximity to an airport)
What is a multivariate correlational design?
More than two measured variables and their correlation
How does the number of variables measured by researchers impact the control over confounds within a study?
With the presence of more variables, the researchers are able to cross examine the data and determine which of the variables truly had the influence ``
How is one way to improve internal validity?
Through measuring more variables
What is the defining characteristic of a regression analysis?
Regressions are used to decide whether variables can predict a separate variable (SATs to college GPA)
What is the function of Beta (standardized coefficient?
Beta is used to quantify the standard deviation between variables (average change that the variables cause)
What is the function of regression coefficient b (non-standardized coefficient)?
Coefficient b is used to quantify the single unit of change the variables cause (this is why it cannot be generalized for all b’s)
What is the purpose of including all of the variables within the regression equation (even if they weren’t significant)?
In order to control for confounds, therefore their minimal (possible) impact is included
What are some ways to tell that a multiple regression has been used in a study?
- “Controlled for”
- “Taking into account”
- “Corrected for”
- “Adjusting for”
What is the most important limitation to a multiple regression?
Alone, an MR study does not establish causation
Why can’t an MR study establish causation?
An MR study is unable to establish temporal precedence, a longitudinal design is also required
Define parsimony?
Choosing to use the best explanation that requires the fewest exceptions or qualifications (apple falls down because the atmosphere around causes the apple to fall, instead of the apple falls because it wants to be back to its most natural place)
What is a mediation model?
Proposed variables that influence one another, typically used in clinical settings (many reasons for anxiety/depression)
What is mediation used for?
To prove temporal precedence within a regression study (SAT’s predict college GPAs, multiple variables contribute to this such as ability to do tasks under a stressful time requirement, logic, and comprehension skills)
How do we establish causation within a study?
Covariance (r), temporal precedence, and internal validity
How do we establish temporal precedence within a correlational/regression study?
Longitudinal design
What are the types of longitudinal results?
cross-sectional correlation, autocorrelation, andcross-lag correlation
Define cross-sectional correlation.
The association is measured between two variables at the same point in time (violence in children and videogame preference at 14)
Define autocorrelation.
The association between the same variable is measured at different points in time (Violence in children at 4 and then at 14)
Define cross-lag correlation.
The association between all the variables at once (violence and videogame preference in 4 and 14 y/o)
Why do a correlational study instead of an experiment?
It opens up the opportunity for studies that may be unethical/unable to be manipulated (smoking studies/videogame preference)
