Causality and Confounds
The way we assert one thing happened after the other.
Temporal Precedence
This type of design has a one-way (one IV) and factorial (multiple IV's) on subjects through random assignment (experimental) or matched comparison (quasi-experimental)
Between Subject Design
Why can't you use a regression for a within-person design?
cross-sectional correlation self-reports
How are conclusions linked to designs and why should our questions inform them?
The design helps us to parse out our hypothesis by indicating, manipulating, correlating and controlling for other factors to help support our theory.
The design must align with the aspects of the question and utilize the appropriate measures, groups, and analysis, as well as the relevant constructs to establish validity in your findings
A kind of relationship that we use to assert there is no third variable or alternative hypothesis.
A non-spurious relationship.
Factorial design levels are...
In between subject we expose subjects to different levels (treatment + meds etc..) based on our hypothesis and measure against treatment no meds.
Ergodicity is known as.
The idea that what we observe in a person would similarly be observed in a group.
Correlational designs, which are usually between person and utilize correlations and regressions, has what limitations?
Explain the ideas of Power and Pre-registration in research design.
Power is our ability to detect results and is determined by the effect we are looking for and sample sizes.
Pre-registration is when we submit our prospective analysis and hypothesis prior to submitting data. This helps with trustworthiness in our study and assure we aren't looking for p values in measure we didn't originally intend to study. (P-Hacking / Harking)
Unlike covariates, which influence results, other factors may explain our outcomes besides what we see in our data, these other factors are known as?
Confounds
While both are considered experimental or quasi-experimental, between subject differes from within subject designs in that...
Compares mean of a subject across the study instead of groups.
This design is always longitudinal while the other can be both correlational and longitudal
Within-person is always longitudinal, between person can be both.
What are the four main types of Longitudinal designs?
Cohort – Different people/Different Times – Share common environment or characteristics. TIME is the Panel – Same people over time period. Prospective Longitudinal – Baseline entry to study. Participants selected for exposure or experience can sometimes be assessed pre-event. Ecological momentary assessment (High-intensity daily study). Measures Daily fluctuations.
False positives have two types of errors, can you name and explain them? How does low-power influence them?
Type I error: Findings support Hypothesis but incorrect.
Type II error: Findings did not suppoer hypothesis but incorrect.
Low power makes for a coin-flip in possibly getting these errors and get significance.
What are three typical types of confounds?
Person-related, Procedural (situational), and Operational (measurement).
Although within subjects helps handle person confounds, as people are their own controls, what are some issues within-subject designs can have?
Issues of order-effects, inter-treatment contamination, practice-effects, sensitization and carry-over.
What are some advantages and disadvantages of within subject?
+ Requires longitudinal data, Good for state variables (affect). - Analysis more complex
LoCS refers to this type of research that often uses betwen and within desgins and can implicate directionality in a study.
Longitudinal correlational self-report.
What is HARKing? also Explain P-hacking and how P-curve analysis helps to discover it?
Harking = Hypothesis After Results Known.
P-hacking is manipulating your data in ways to obtain P<.05 (controlling for removing etc..) and implies your data is true when your findings were different from hypothesized.
P-curve allows us to plot p-values across a body of research to see if the values hover around .001 (expectation of true observation) or .05 (possible p-hacking).
Some ways we address confounds are?
Experimental designs (helps eliminate person confounds), careful measurement, and statistical control for confounds.
What is a primary difference for within and between persons compared to subjects?
Persons is correlational and examins charecteristics in groups (between) or across measures of a person (within).
What are some advantages and disadvantages of between person?
+ Can use cross-sectional data, Good for trait variables (big five ex), easier and more straightforward - may not accurately characterize relationships within people.
What are Experimental (TeRAD) and Quasi-Experimental (QuaD) designs? And what differentiates them?
TeRAD: True Experiment Random Assignment Design - used in a controlled setting with random assignment.
Quad: Used in real world settings to test for causal relationships.
Generalizability, conclusion validity, and external validity refer to what?
Generalizability - ability to extrapolate our results across larger more general populations.
Conclusion validitiy - being able to draw a fair and legitimate conclusion that is stays close to our data.
External validitiy - would our results be valid outside the experiment in the real world?