2.3 Analyzing Findings

Analyzing Research Findings

Correlation Coefficients in Psychological Research

A correlation coefficient quantifies the strength and direction of the relationship between two variables. For example, a researcher might measure the correlation between stress levels and sleep quality.

Correlation coefficients range from -1 to +1:

• +1 = a perfect positive correlation (as one variable increases, the other increases proportionally)
• 0 = no correlation (no consistent relationship between the variables)
• -1 = a perfect negative correlation (as one variable increases, the other decreases proportionally)

The closer the coefficient is to -1 or +1, the stronger the relationship. An $r = 0.85$ between study time and test scores would suggest a strong positive correlation, while an $r = -0.30$ between exercise and anxiety might suggest a weak-to-moderate negative one.

Statistical significance tells you whether the observed correlation is likely real or just due to chance. It does not tell you how strong the relationship is.

• The standard threshold is $p < 0.05$, meaning there's less than a 5% probability the result happened by chance.
• A weak correlation like $r = 0.2$ can still be statistically significant if the sample is large enough. Significance and strength are separate questions.

Effect size measures the magnitude of a relationship or difference, complementing significance. Cohen's $d$, for instance, is commonly used to express how large the difference is between two groups.

Correlation vs. Causation

This distinction comes up constantly in psychology, and it's one of the easiest places to lose points on an exam.

Correlation describes a relationship between two variables without implying that one causes the other. The classic example: ice cream sales and drowning rates are positively correlated, but ice cream doesn't cause drowning. A third variable, summer weather, drives both.

Causation means one variable directly produces a change in another. Smoking causes lung cancer, for instance. But establishing causation requires more than just observing a correlation. You typically need:

1. A controlled experiment where the researcher manipulates one variable
2. Random assignment to rule out pre-existing differences between groups
3. Evidence that no third variable better explains the result

Carrying a lighter correlates with lung cancer rates, but the third variable (smoking) is the actual cause. Always ask yourself: could something else explain this relationship?

Biases in Variable Relationships

• Confirmation bias leads researchers to focus on evidence that supports their existing beliefs while ignoring contradictory data. A researcher who expects a certain outcome might unconsciously give more weight to studies that confirm their hypothesis and dismiss ones that don't.
• Illusory correlation is perceiving a relationship between variables when none actually exists. People who believe in astrology, for example, may notice personality traits that "match" someone's sign while overlooking all the times the traits don't match.
• Third variable problem occurs when an unmeasured variable is actually driving the relationship between two observed variables. The ice cream and drowning example above is a textbook case: summer weather is the hidden third variable causing both to rise together.

Advanced Analysis Techniques

• Meta-analysis combines results from multiple studies on the same question to produce a more reliable overall conclusion. It's especially useful when individual studies have small samples or mixed results.
• Regression analysis examines how multiple variables relate to an outcome, allowing researchers to predict scores on one variable based on one or more predictors.
• Replication means repeating a study to see if the findings hold up. If results can't be replicated across different samples and contexts, the original findings may not be reliable.

Experimental Design and Interpretation

Random Sampling and Group Assignment

These are two different techniques that serve different purposes. Mixing them up is a common mistake.

Random sampling is how you select participants from a population. Every member of the population has an equal chance of being chosen (e.g., using a random number generator to pick names from a list). This increases external validity, meaning the results are more likely to generalize to the broader population.

Random assignment is how you place already-selected participants into experimental groups. Each participant has an equal chance of ending up in any condition (e.g., a coin flip determines who gets the treatment vs. the placebo). This increases internal validity by making the groups roughly equivalent at the start, which reduces the influence of confounding variables.

Sources of Bias in Experiments

• Demand characteristics occur when participants figure out (or think they've figured out) what the study is testing and change their behavior accordingly. Someone in an anxiety study might act more anxious because they assume that's what the researcher is looking for.
• Experimenter bias happens when a researcher's expectations subtly influence participants or the interpretation of data. Even small cues like tone of voice or facial expressions can push participants toward certain responses. Double-blind procedures, where neither the participant nor the researcher knows who's in which condition, help prevent this.
• Placebo effect refers to improvement that occurs simply because a participant believes they're receiving an effective treatment. A person given a sugar pill might report less pain because they expect the "medication" to work. This is why control groups receiving placebos are so important in treatment studies.

Independent and Dependent Variables

The independent variable (IV) is what the researcher manipulates. It's the presumed cause. For example, in a drug trial, the IV might be medication dosage: one group gets the drug, another gets a placebo.

The dependent variable (DV) is what the researcher measures. It's the presumed effect. In that same drug trial, the DV would be symptom severity after treatment.

A simple way to keep them straight: the DV depends on the IV. The researcher changes the independent variable and then measures whether the dependent variable changed as a result.