Correlational Research

Introduction

Describing people's behavior is often the starting point for behavioral research. Correlational research steps in where descriptive research leaves off by aiming to uncover potential patterns of relationships between variables.

For example, in the previous chapter, we talked about measuring anxiety with the GAD-7. In a correlational study, researchers might move beyond describing anxiety to investigate how anxiety is associated with other variables such as depression, trauma, sleep, or self-esteem. Are anxious people more likely to be depressed? Are married people less likely to be anxious than unmarried people? Do people become less anxious as they get older? These are questions for correlational research.

Understanding associations, or patterns of relationships between variables, is what moves behavioral research from description to prediction. If a researcher establishes that variable A is related to variable B, then it is possible to predict values of variable B simply by knowing values of variable A. An example comes from the survey of mental health we examined in Chapter 3. The survey, conducted by the National Institute of Mental Health, reported that younger people experienced mental illness at a higher rate than older people. This means that older people have a lower risk of developing mental illness than younger people. The ability to predict risk is a big step forward in understanding.

In this chapter, we will learn how to analyze relationships between variables and discover meaningful patterns in behavioral data. In Module 5.1, we will explore the fundamentals of correlation, including how to measure, visualize, and interpret relationships between continuous variables like anxiety and depression. We will learn to distinguish between positive and negative correlations, understand the strength of a correlation, and create correlation matrices that reveal patterns across multiple variables.

In Module 5.2, we will expand our analytical toolkit by examining different types of associations. We will discover how to analyze associations between categorical variables (like gender or employment status) and continuous variables (like depression scores), using techniques such as t-tests and chi-square.

Module 5.3 offers a hands-on research project that applies correlational methods to the Heinz dilemma we encountered in Chapter 3. Using Moral Foundations Theory, you will develop and test hypotheses about how people's moral intuitions predict their ethical judgments. This guided project will walk you through each step of correlational research, from forming hypotheses to creating a study on Qualtrics, to analyzing and reporting the results.

Finally, Module 5.4 empowers you to design and conduct your own correlational study. You will develop a research question, create a survey to test the relationship between variables, and collect and analyze your own data. This independent project will allow you to apply everything you have learned so far and give you the chance to make your own discoveries about human behavior.

By the end of this chapter, you will understand how correlational research helps scientists move beyond simple description to identify meaningful patterns and make predictions about human behavior. You will also have practical experience conducting correlational studies, an essential skill for anyone interested in understanding the complex relationships that shape how people think, feel, and act.

Chapter Outline

In Chapter 3, we learned about descriptive research and the power of observing people's behavior. With systematic observation, researchers can not only describe what people do but also identify patterns, trends, and associations. An association occurs when changes in one variable correspond to changes in another variable in a systematic way. Here's an example from everyday life.

Many years ago, some psychology professors at big schools with competitive football teams observed something interesting: on Mondays after the football team won (the games were always played on Saturdays in those days), lots of students attended class in clothing that displayed the school's name or logo. On Monday's after the football team lost, however, school-affiliated clothing appeared less often. Curious to know if these observations were reliable, and if so, how strong the association between winning and wearing the school's clothing was, the professor's decided to conduct a study (Cialdini et al., 1976). The study is now a classic in the field of social psychology, showing how people try to affiliate with successful teams or people in order to boost self-esteem. For our purposes, it's also an introduction to correlational research.

Correlational research examines whether two variables are related to each other and quantifies their association with statistics. Instead of manipulating or controlling variables like in an experiment, researchers simply measure both variables as they naturally occur and look for patterns. To understand how correlational research works, we will examine a real dataset several times throughout this chapter. The data comes from over 500 participants on Connect and is actually the same dataset you used in the previous chapter to understand the basics of measurement. Now, however, we will explore the relationships between variables rather than just examining them individually.

As a reminder, the study was created in Qualtrics and included several validated measures. You can examine the survey by downloading the "RITC_SURVEY_CH05_ClinicalStudy.qsf" file from the "Chapter 5 – Correlational Research" folder on the OSF project page: https://osf.io/a8kev/. Once you have this file, upload it into Qualtrics or Engage to view the survey.

Within the survey, participants completed the Generalized Anxiety Disorder scale (GAD-7), as well as the Patient Health Questionnaire (PHQ-9), a widely-used measure of depression (Kroenke et al., 2001). Then, they completed a measure of sleep quality and experiences of trauma. In the last chapter, we calculated total scores for each of these variables. You can either use the data file you worked with previously or download a new data file from the Chapter 5 project folder, in which these total scores have been calculated for you. The file name is: RITC_DATA_CH05_ClinicalStudy.sav.

Understanding Correlations

The correlation between variables is expressed as a statistic, called a correlation coefficient. Pearson's r is the most common correlation coefficient, and it provides information about both the direction and strength of the relationship between two variables.

Positive Relationships

Positive correlations occur when an increase in one variable predicts an increase in another variable. Lots of variables are positively correlated. People with more education tend to earn more money (e.g., Day & Neuberger, 2002). People who spend more time on social media tend to feel more socially isolated (e.g., Primack et al., 2017). And people who have high levels of anxiety tend to also have higher levels of depression (e.g., Löwe et al., 2008; Spitzer et al., 2006). As these examples show, in positive correlations, people who score high on one measure also tend to score high on the other. This kind of relationship is often visually depicted with a scatterplot where the trend line ascends from left to right (Figure 5.1) and is mathematically represented by a positive correlation coefficient.

Correlations, Trends, and Prediction

When examining correlations, the predictions researchers make are not perfect for every person. You can see this by examining the scatterplot in Figure 5.2. Notice that some people have depression scores above ten but anxiety scores of zero. For these people, high depression scores do not predict high anxiety scores. Nevertheless, across the entire sample, people with high anxiety scores generally have higher depression scores and vice versa.

Clinical psychologists know that anxiety and depression usually have a strong ability to predict one another because the correlation between these variables is one of the most fundamental findings within the field (e.g., Kalin, 2020; Spitzer et al., 2006). This strong association helps explain why many mental health treatments target both anxiety and depression simultaneously.

Negative Relationships

Not all correlations are positive, some are negative. A negative correlation indicates that as scores on one variable increase scores on the other variable tend to decrease. This is also sometimes called an inverse relationship.

A negative correlation is how we would describe the relationship between anxiety and self-esteem. As people's anxiety increases their self-esteem generally decreases (e.g., Löwe et al., 2008). Thus, these two variables are negatively correlated.

Other examples of negative correlations abound. For example, as people's time spent multitasking increases, their satisfaction with what they are doing tends to decrease (Mark et al., 2008; Mark et al., 2016). The less time people spend sleeping, the more health problems they tend to encounter (Luyster et al., 2012). And, as people's sense of control over their lives decreases, the frequency of depression tends to increase. When a negative correlation is graphed, the result is a line that moves down and to the right as in Figure 5.3.

When a Correlation Counts: Magnitude and Statistical Significance

The Magnitude of a Correlation

While the sign of a correlation coefficient tells us how two variables are related, it does not say anything about the strength of the relationship. For that information, we must examine the numerical value of the Pearson's r statistic. The closer the value is to 1 or -1, the stronger the relationship. The closer the value is to zero, the weaker the relationship. And, if the correlation is close to zero there is no relationship between the variables (Figure 5.4).

Across the behavioral sciences, researchers share conventions for interpreting the size of a correlation coefficient. These conventions come from the work of Jacob Cohen, a prominent figure in the field of psychology and statistics (1992). The conventions apply to the size of a correlation regardless of its direction.

A correlation between 0.1 and 0.3 is considered small, indicating a weak association between the variables. A coefficient above 0.3 and below 0.5 is considered moderate, suggesting a more substantial but still modest relationship. Finally, a correlation of 0.5 or higher is considered large, indicating a strong relationship between the variables. From the examples we have seen, the correlation between anxiety and depression would be considered large and the correlation between depression and age would be considered small. Table 5.1 presents other examples of small, moderate, and large correlations.

Table 5.1. Guidelines for the strength of correlation coefficients. These conventions are the same for positive and negative values.

Statistical Significance

A correlation that is close to zero indicates no relationship between variables. Yet when behavioral scientists calculate the correlation between any two variables, they rarely get exactly zero. Even if two variables should not be related—say, shoe size and empathy—there will probably be some small correlation, maybe .03 or -.04.

The question then becomes: when is a correlation large enough to be meaningful rather than just random noise? The answer comes from statistical significance. Researchers use tests of statistical significance to determine if a correlation is larger than what would occur by chance. The result of these tests is expressed as a probability or "p-value".

By convention, if p is less than .05 the correlation is statistically significant. When p < .05, behavioral scientists have good reason to believe the correlation represents a real relationship rather than random variation in the data. A p-value below .05 means that if there were truly no relationship between the variables, a correlation as large as what is observed (or larger) would be found less than 5% of the time by chance alone.

In the data you examined earlier, all the correlations were statistically significant. The strong positive correlation between anxiety and depression (r = .82, p < .05) and the negative correlations between age and depression (r = -.20, p < .05) and age and anxiety (r = -.23, p < .05) all represent reliable patterns in the data, rather than chance findings.

The logic of statistical significance applies to all the statistical tests you will encounter in this book. Whether looking at correlations or any other statistical test, behavioral scientists use p < .05 as a guideline for determining whether the results are meaningful or more likely due to chance.

📊

Research Activity 5.3: Examining a Correlation Matrix

When there are multiple variables in a study, researchers often want to examine how all the variables relate to each other. Rather than calculating correlations one at a time, it is possible to generate a correlation matrix—a table that shows all possible correlations at once.

To create a correlation matrix in SPSS, follow the steps in HOW TO Box 5.1 or watch the video associated with this activity: https://bit.ly/Ch5_cm. The process is similar to what you did earlier, but with one additional step. Instead of placing two variables into the "Variables" box in the correlation analysis, move all the variables of interest into the box. The resulting output will show every possible correlation between the variables in one table.

Figure 5.5 shows a correlation matrix with seven variables from the clinical dataset. The variables in the matrix include depression, anxiety, sleep quality, trauma, age, income, and education. Each cell in the matrix shows the correlation between the variables in that row and column. For instance, to find the correlation between depression and anxiety, look at where the depression row intersects with the anxiety column. The value there is r = .82.

A correlation matrix has several key features. First, along the diagonal line running from top-left to bottom-right, we see all 1's—these represent each variable's correlation with itself.

Second, above and below the diagonal line, the matrix is symmetrical. This means the correlation between any two variables appears twice. When reading the table, we can ignore everything below the diagonal. Third, the "Sig. (2-tailed) value" is the p value for each correlation. Asterisks mark statistically significant correlations, with one asterisk indicating p < .05 and two asterisks indicating p < .01. Finally, the "N" rows show the sample size for each correlation.

Looking at the matrix reveals patterns in how these variables correlate. For example, depression shows a strong positive correlation with both anxiety (r = .82) and sleep problems (r = .71), a moderate positive correlation with trauma (r = .54), and a small negative correlation with both age (r = -.20) and income (r = -.16). The correlation between disordered sleep and education is not significant.

Reporting the Results of a Correlation Analysis

Writing about statistical results is an important part of behavioral research. It is used not only to clearly communicate the results of a study to others, but often it helps clarify a researcher's own thinking about what was found. A good write-up presents the results in clear, non-technical language that any reader can understand, while also including the essential statistical information. Let's look at an example.

To describe the correlation between anxiety and depression you might write:

"I conducted a Pearson correlation analysis to examine the relationship between anxiety and depression. I found a significant positive correlation between anxiety (as measured by the GAD-7) and depression (as measured by the PHQ-9), r(524) = .82, p < .05. This result suggests that anxiety and depression are strongly related. Specifically, the results suggest that people who experience higher levels of anxiety also tend to experience higher levels of depression."

In this write-up, the correlation coefficient (r = .82) tells readers how strong the relationship is and in what direction. The number in parentheses (524) represents the degrees of freedom, which is related to the sample size (sample size [N] minus 2). The p-value (p < .05) indicates the relationship is statistically significant and larger than what would be expected by chance. Together, these elements give readers the information they need. Plus, the plain language interpretation helps everyone understand what the results mean in practical terms.

Types of Associations

Many behavioral studies examine how different variables are associated with one another. In conducting these studies, researchers are interested in whether changes in one variable correspond to changes in another variable. As we saw in the last chapter, however, not all variables are measured on the same scale. Variables measured on different scales produce different kinds of associations.

Recall that the previous chapter described four scales of measurement: nominal, ordinal, interval, and ratio (remember the acronym 'measurement NOIR'). For practical purposes, these measurement scales can be simplified into two broad categories: categorical and continuous. So far, this chapter has focused on associations between continuous variables like anxiety and depression. However, behavioral scientists often want to understand relationships between variables that are not continuous. For example, is anxiety (continuous) related to gender (not continuous)? Examining this kind of relationship requires a different statistical tool than the ones we have seen so far. When examining the relationship between variables that are not both continuous, researchers must use different statistical tools.

When examining associations, we can treat both interval and ratio scales as continuous variables because they exist on a continuum and are analyzed the same way. We can also treat ordinal variables, like Likert scales, as continuous variables (even though there are specialized techniques for ordinal data that we do not cover here). Finally, we can treat nominal variables as categorical data. This yields two main types of variables—categorical and continuous—that can be associated with each other in three possible combinations (Figure 5.6).

Each type of relationship requires a different analytical approach. The type of relationship also determines how researchers visualize the results. Let's examine each type of relationship using examples from the anxiety dataset.

Associations between Categorical and Continuous Variables

Many research questions examine how different groups of people differ on a continuous measure. A school administrator, for instance, might want to know if students at private schools have higher test scores than students at public schools. In this case, school type (public vs. private) is categorical while test scores are continuous. Public health researchers might investigate whether people living in cities report higher stress than those in rural areas. Medical researchers might examine whether participants in exercise programs show lower blood pressure than people not in such programs. Or developmental psychologists might want to know whether first-born children score differently on personality traits than later-born children. In each of these cases, the researchers want to know if one group of people scores differently on a measure than another group.

To analyze these kinds of relationships, researchers compare the average scores for each group. For example, in the anxiety dataset, we can examine whether the average depression score is higher for one gender group than another.

Conducting a t-test

Just as we used Pearson's r to describe the relationship between continuous variables, researchers use something called a t-test to determine if differences between group averages are meaningful. The t-test assesses whether the difference between two groups is larger than what would be expected by chance alone. Like with correlations, a p < .05 is the criterion for statistical significance.

HOW TO Box 5.2 describes how to conduct the t-test examining whether men's and women's depression scores differ on average. You can follow the instructions in the box or watch the video for this project: https://bit.ly/Ch5_CtT.

The t-test shows a significant difference between men's and women's depression scores (t = 2.3, p < .05), meaning the difference is larger than what would be expected by chance (Figure 5.8). This significant test confirms that there is a real, reliable relationship between gender and depression in this sample.

Just as with correlations, researchers communicate the findings of t-tests in a scientific report. An example of how to do this might read:

"I conducted an independent samples t-test to examine differences in depression between men (M = 6.10, SD = 5.31) and women (M = 7.27, SD = 6.49). There was a significant difference, t(525) = 2.38, p < .05, suggesting that gender is associated with depression. More specifically, my results indicated that women tend to experience higher levels of depression than men, even though the effect size (Cohen's d = 0.21) suggests the difference is small in magnitude."

If you are wondering where that last bit about effect sizes came from and why d = .21 is considered small, continue to the next section.

Making Sense of Effect Sizes

Just as researchers have guidelines for interpreting the size of correlations, they have guidelines for interpreting the size of differences between group means. The most common statistic for this is Cohen's d, which measures the difference between two group means relative to the standard deviation of both groups.

A Cohen's d around 0.2 represents a small difference between groups. Values around 0.4 represent medium-sized differences. And values of 0.6 or larger represent large differences. These guidelines help researchers communicate not just whether group differences are statistically significant, but how meaningful they are in practical terms. Table 5.2 contains examples of research findings with different effect sizes.

In our sample project, we found a Cohen's d of 0.2, indicating a small but reliable difference between men's and women's depression scores. This aligns with the survey from Chapter 3, where women showed somewhat higher rates of mental illness than men. While the difference is small, the consistency of this pattern across different studies—from large national surveys down to our sample of 500 participants—suggests the gender difference in depression is a reliable phenomenon, even if its effects are modest.

Table 5.2. Conventions for interpreting Cohen's d effect sizes.

Overall, the key to understanding categorical-continuous relationships is recognizing that they are still just a pattern of association. However, instead of looking at how two continuous variables move together (like anxiety and depression), researchers compare the average scores of different groups. The statistical tools change—from correlation coefficients to t-tests—but the goal remains the same: examining whether scores on one variable help predict scores on another.

Associations between Two Categorical Variables

Many questions in behavioral research involve relationships between categorical variables. For example, researchers might want to know whether employed people are more likely to register as voters than unemployed people. In this case, both variables—employment status (employed vs. unemployed) and voter registration (registered vs. not registered)—are categorical. Similarly, researchers might investigate whether college graduates are more likely to own homes than non-graduates, or whether people living in cities are more likely to use public transportation than those living in rural areas.

When examining relationships between categorical variables, researchers compare percentages between groups: what percentage of employed people are registered to vote vs what percentage of unemployed people are registered to vote? To examine whether percentages differ between groups, researchers use a statistic called a chi-square (χ²). The idea behind the chi-square is the same as Pearson's r and the t-test: we want to know if the p-value associated with the statistic is smaller than .05. If so, there is an association between the variables.

To report the results of a chi-square analysis, you might write:

"I conducted a chi-square test to examine if there was a difference in the percentage of men and women who scored above a 20 on the PHQ-9. There was a significant difference, χ² (2) = 10.71, p < .05, suggesting that gender is associated with severe depression. More specifically, my results indicated that women tend to experience severe depression at a higher rate than men."

Guided Project: Moral Foundations and the Heinz Dilemma

Throughout this chapter, we have explored how researchers examine the associations between variables. Now, we will put this knowledge to use in a guided project.

Like the project we completed together in Chapter 3, this project will involve the Heinz dilemma. Instead of describing people's moral decisions, however, we will examine what might predict those decisions. Specifically, we will use something called Moral Foundations Theory (e.g., Graham et al., 2013) to investigate whether differences in people's moral intuitions help explain their judgments in the Heinz dilemma. All the study materials, the data file, and the instructions for what to do will be provided to you.

The project will give you hands-on experience with the key elements of correlational research: developing theoretically driven hypotheses, analyzing relationships between variables, and interpreting results. The accompanying video for this project provides a step-by-step guide for what you need to do: https://bit.ly/Ch5hcs.

What We are Studying: Project Goals and Big Questions

This project consists of three tasks. First, you will generate research hypotheses. This will involve reading about Moral Foundations Theory and developing two specific hypotheses about how moral foundations might predict reactions to the Heinz dilemma. For each hypothesis, you will write approximately one paragraph explaining your rationale based on the Moral Foundations theory.

Second, you will analyze the data. After examining the Qualtrics survey, you will download the SPSS data file and calculate scores for the five moral foundations subscales. You will then create a correlation matrix showing relationships between all five moral foundations and the moral acceptability ratings. Next, you will conduct t-tests to examine whether moral foundation scores differ between people who think Heinz should have stolen the drug and those who think he should not have. For your hypothesized relationships, you will create appropriate figures, including a bar graph comparing moral foundation scores between groups.

Finally, you will write up the results. Your write-up should focus on the specific hypotheses you generated, following the examples from earlier in the chapter that show how to report statistical analyses.

Part 1: Frame Your Hypotheses

In Chapter 3, we conducted a descriptive study examining how people respond to the Heinz dilemma. We found that when people were asked whether Heinz should steal the drug, a slight majority said he should not. We also found that when people were asked how morally acceptable stealing would be, most found it acceptable.

In this project, we will explore what might predict people's responses. To answer this question, we will draw upon Moral Foundations Theory. Just as the Big Five personality traits provide a framework for understanding personality, Moral Foundations Theory suggests that moral judgments can be understood through five basic moral intuitions that are described in Box 5.4.

Brief Background Reading

This project begins where all research begins: looking at existing theory. Take 20-30 minutes to read about Moral Foundations Theory. You can start at the www.moralfoundations.org website. Then, you can explore the peer reviewed literature on Google Scholar (e.g. Graham et al., 2013). During your reading, pay attention to how this theory has been used to understand people's moral judgments in different situations.

Develop Your Hypotheses

After familiarizing yourself with the Moral Foundations Theory, take a moment to develop two of your own hypotheses. Looking at the five moral foundations, which ones do you think might predict whether someone finds stealing the drug acceptable or unacceptable? Why?

Write down your predictions and your reasoning in your portfolio. Specifically, one hypothesis should relate to the question of why people find Heinz's decision morally acceptable or unacceptable. The second hypothesis should relate to whether Heinz should have stolen the drug. For example, are people who say that Heinz should steal the drug more likely to have higher Authority scores compared to those who think he shouldn't steal the drug?

Part 2: Design, Materials, and Methods

Now that you have hypotheses, let's examine how to test them. The first step is selecting the measures to assess each construct.

Moral Foundations Questionnaire

To measure moral foundations, we will use the Moral Foundations Questionnaire (MFQ-30), which assesses how strongly people endorse each of the five moral foundations (Graham et al., 2008). The MFQ-30 contains 30 items total, with six items measuring each foundation. For instance, an item that measures the Care/Harm foundation asks people to rate how relevant "whether or not someone suffered emotionally" is when deciding if something is right or wrong. All items are answered on a 1 to 5 scale, although the answer labels vary across the measure.

Heinz Dilemma

We will present participants with the same moral dilemma and follow up questions used in Chapter 3. The questions asked whether Heinz should steal the drug (yes/no) and how morally acceptable stealing would be (rated from 1-7).

Data Quality

The Moral Foundations Questionnaire (MFQ-30) includes a few items that check to see if participants are paying attention. For example, one item asks participants how important "whether or not someone was good at math" is when deciding between right and wrong. To the items embedded within the MFQ, we added an additional attention check. Chapters 10, 11, and 12 cover data quality in-depth, but as an optional exercise, you can identify which participants failed the attention checks and remove them from the data analyses after performing the steps below.

Accessing study materials

To see how the measures were implemented online, download the Qualtrics survey file from the OSF project page: https://osf.io/a8kev/. Find the folder named "Ch. 5 – Correlational Research" and download the "RITC_SURVEY_CH05_HeinzDilemma.qsf" file. Import this file to your Qualtrics or Engage account and explore its structure.

Notice how the survey is organized into blocks, how it uses matrix-style questions to present the MFQ-30 items, and how it randomly determines the order people receive the MFQ-30 and Heinz dilemma in (you can see this within the survey flow). The instructional video for this assignment walks you through the key features of the survey design in more detail.

Data collection

After programming the survey, we gathered data from 200 Connect participants. Each person was paid $1.00 for their time, and the study took about 7 minutes to complete. Once the study was launched, data collection completed in under an hour.

Part 3: Analyze the Data

To analyze the data, download the SPSS file from the OSF page. In the "Ch. 5 – Correlational Research" folder find the file named "RITC_DATA_CH05_HeinzDilemma.sav" and download it.

Once you have the file open, calculate the scores for each moral foundation subscale. HOW TO Box 5.5 provides instructions or you can watch the video for this project. The MFQ-30 consists of five moral foundations with six items each. You need to average the items to create a single score for each foundation.

Once you have created a score for each subscale, create a correlation matrix to show the correlations between all five moral foundation subscales and the moral acceptability judgments. Then, conduct t-tests to see whether people who thought it was acceptable for Heinz to steal the drug differ from those who thought it was unacceptable. These analyses will tell you whether your hypotheses about the moral foundations and the Heinz dilemma were supported.

Create a Figure

After the t-tests, create a bar graph that compares moral foundation scores between those who thought Heinz should steal the drug and those who thought he should not. You can use either HOW TO Box 5.3 or the video for this assignment to guide you through this process. Make sure you put the yes/no groups on the x-axis and the moral foundation score on the y-axis. Also make sure your graph includes error bars (± 1 standard error). Label both axes and give the graph a title. You will include this figure with the write-up of your results.

Part 4: Interpret the Findings

How do the results compare to your hypotheses? Were your predictions supported? What surprised you about the findings? Why do you think some moral foundations predicted reactions to the dilemma while others didn't?

Write up the results relating to your hypotheses about moral foundations and acceptability ratings using the style below:

"I conducted a Pearson correlation analysis to examine the relationship between [your predicted moral foundation] and judgments of moral acceptability in the Heinz dilemma. There was a [significant/non-significant] [positive/negative] correlation between [foundation name] and moral acceptability ratings, r(198) = [value], p = [value]."

Add one sentence explaining what your result means in plain language. Include your figure and a brief statement about whether your hypotheses were supported.

Group Differences

For your hypothesis about moral foundations and people's yes/no decisions, write up your findings with the following structure:

"I conducted an independent samples t-test to compare [foundation name] scores between participants who said Heinz should steal the drug and those who said he should not. There was a [significant/non-significant] difference in scores for yes (M = [value], SD = [value]) and no (M = [value], SD = [value]) groups; t(198) = [value], p = [value]."

Add one sentence explaining what this means in plain language.

Now that you have worked through an example, you are ready to investigate your own correlational research question.

Throughout this book, we have explored several different domains of research. You have measured personality traits, examined clinical variables like anxiety and depression, and investigated stages of moral development. Now it is time to use what you have learned to investigate a question that interests you.

Your correlational study could build on any of the measures explored previously. For instance, you might wonder how personality traits relate to mental health: are more conscientious people less likely to experience anxiety? Or you might be curious about how moral foundations connect to other aspects of behavior: do people who score high on the care/harm foundation show more empathy in everyday situations?

You could also venture into new territory. Many students are curious about patterns they have observed in their own lives or questions about human behavior that intrigue them. For example: Do people who spend more time on social media report feeling more socially connected or more isolated? Are students who exercise regularly less stressed during exam periods? Do people who maintain structured daily routines sleep better? Is procrastination related to perfectionism?

You have experience finding validated measurement tools, from the TIPI for assessing personality to the GAD-7 for anxiety and the Moral Foundations Questionnaire for moral intuitions. You can use these measures, find new measures in the instrument databases we discussed, or create your own measure following the process outlined in Chapter 4 for working with AI. Either way, this project presents an opportunity to pull together many of the things you have learned to investigate a question of your choosing.

Step 1: Craft Your Question and Study Design

Using Qualtrics or Engage, design a short survey to collect data for a correlational research project. Keep it simple: measuring two or three variables is plenty. The best way to begin is to take the Moral Foundations study in Qualtrics and modify it. Include clear instructions for participants and organize your measurements into blocks.

The first block should include instructions and a welcome message. After that, you should measure one variable per block, placing all the items that are part of your measurement instrument into their own block. The last block should include basic demographic questions.

Step 2: Collect Data

For this project, you should aim to collect data from at least 100 participants. These could be students in your school (if your instructor sets up a class data collection system through a platform like Sona), friends and family (using the anonymous survey link from Qualtrics or Engage), or CloudResearch Connect participants.

Step 3: Analyze Your Results

Once you have the data, analyze it using the statistical tools you learned about in this chapter. For relationships between continuous variables, use correlations. For comparing groups on continuous measures, use t-tests. It is less common to examine relationships between categorical variables for a first attempt at a correlational study, but if this fits the question you have chosen, use chi-square tests. Create appropriate visualizations using the techniques you practiced in the chapter.

Summary

In this chapter, we explored correlational research. We learned how to analyze correlations, interpret their strength, and understand when relationships are statistically significant. We also saw how behavioral scientists assess relationships between different types of variables: continuous-continuous associations are measured with Pearson's r, categorical-continuous associations with t-tests, and categorical-categorical associations with a chi-square.

The hands-on projects have given you practical experience with correlational analyses. By exploring relationships between anxiety, depression, and demographic variables, you had the chance to work with a real dataset. The Heinz dilemma project further demonstrated how psychological theories like Moral Foundations Theory can explain individual differences in moral judgments.

The goal of correlational research, as should be clear by now, is to identify patterns and make predictions. When behavioral scientists find that two variables are correlated, they know that one variable can predict the other. However, correlations alone cannot tell researchers whether one variable causes change in the other.

This limitation brings us to the next chapter, where we will explore how researchers address questions of causality in correlational research. Chapter 6 will introduce you to advanced techniques for investigating cause and effect relationships when experiments are not possible. You will learn about statistical control, multiple regression, and longitudinal designs. These are common methods that help researchers build stronger evidence for causal relationships while still acknowledging the inherent limitations of correlational approaches.

By combining what you have learned about correlational research with these advanced techniques of causal inference, you will develop a more sophisticated understanding of how behavioral scientists investigate associations between variables and draw meaningful conclusions about human behavior.