Research Design: Experiments and Quasi-Experiments

Study Notes

Research Design

Experiments: classic research design where variables are manipulated, allowing for drawing conclusions about cause and effect, and establishing causal relationships between variables.
Quasi-experiments: research design similar to experiments but without full control over variables, used when it's impossible, impractical, or unethical to manipulate the IV.

Surveys

Emphasis on general statements about a sample, with a focus on non-directional relationships between variables.
Types of surveys:
- Face-to-face interviews
- Self-administered questionnaires
- Computer-mediated questionnaires
- Observational surveys
- Archival surveys (looking through archival means)
- Behavioral trace measures (e.g., tweets, Instagram posts)
Type of survey chosen depends on:
- Constructs being operationalized
- Likelihood of reactance
- Likelihood of social desirable responding
- The degree to which participants can accurately respond

Statistics

Descriptive statistics:
- Used to describe or characterize a population
- Specify where the data falls on a relevant scale
- Can be used for making predictions (where the data will fall)
- Can be used for quantifying uncertainty (worse at making predictions the more spread out the data is)
Inferential statistics:
- Used to infer characteristics of a population from a sample
- Used to test specific hypotheses
- Are two samples drawn from the same population?

Levels of Measurement

Nominal: discrete categories, no order
Ordinal: continuous categories, has an order
Interval: continuous rank-ordered scale, where points are meaningful and quantifiable (e.g., temp in Celsius)
Ratio: same as interval, but zero point represents the absence of constructs being measured (e.g., temp in Kelvin)

Measures of Central Tendency

Characterize where the data falls on the scale
The mode: most frequent
The median: the middle
The mean: mathematical center/average

Measures of Dispersion

Characterize how spread out the data is
Allow us to quantify uncertainty in our predictions
The more spread out the data, the worse the measure of central tendency is as a predictor
Range: max - min
Variance: average deviation of scores from the mean in squared units
Standard deviation: the average deviation of scores from the mean in raw score units

Distributions

The way in which our data is spread along a scale when graphed by frequency per scale point
In psych, most variables are normally distributed, mean = mode = median
Distributions can be skewed

Null Hypothesis Significance Testing (NHST)

Assumes a null hypothesis to be true and then assesses the probability of getting an observed result under that assumption
Null hypothesis: the hypothesis that there will be no effect
Used for:
- Are two sample means different, H(0) - the sample means are equal/not different
- Are two interval/ratio variables related, H(0) - the two variables are unrelated - Pearson's correlation coefficient
In practice, NHST relies on the properties of distribution
In each case, we assess the probability of the observed outcome, assuming the null hypothesis (H(0)) is true

Description

Understanding the differences between experiments and quasi-experiments in research design. Learn how to establish causal relationships between variables and the role of random assignment and control in experiments.