Everyday Research Methods

Which brand of car is the most reliable?

How might advertisers operationalize the concept of "reliability" in ways favorable to their own brand?  Credit: BNFWork/Shutterstock

You may have seen these ads on TV: Real people stand in a warehouse and are asked which cars are "most reliable." A Honda driver might predict, "Hondas are the most reliable." But then the cover is whisked off in a flourish: The most reliable car is actually the Chevy! In one recent ad (viewable here), the narrator reports: 

"Based on a recent nationwide survey, Chevy is more reliable than Toyota, Honda, and Ford."

Should statements like these convince you? The website Jalopnik publishes journalism about cars. In this column, a journalist walks readers through how they might think critically about  reliability claims in the Chevy commercials. For example, the journalist writes that the reliability claim was based on a survey.

...the survey was sent in 2018 to owners of 2015 model year cars, and the reporting of the “repairs in the past 12 months” relates to those 2015 cars in their third year of service. Chevy reiterated that “Independent statisticians reviewed the materials and concluded with 95% confidence that Chevy’s percentage of no parts repaired or replaced is better than Toyota, Honda, Ford or 23 other brands.”

The journalist adds:

Anyone who has studied research and statistics will tell you that it’s not difficult to form the conclusions you want to based on your use of what is called an “operational definition.” For this particular study, Ipsos is operationally defining reliability as the repairs a car has had within a 12 month period once the car has reached three years old.

a) How did Chevy operationalize "reliability" in their study? 

b) What are some alternative ways to operationalize a car brand's "reliability?"

c) Can you think of downsides to the operationalization that Chevy used? 

The journalist pointed out that, other than the operationalization of reliability, the survey's number of respondents might matter. To wit:

Setting aside the fact that self-reported surveys can be flawed, fewer than 49,000 respondents completed them out of almost 840,000. What Chevrolet did not provide is a breakdown of those respondents by brand, because if we were looking at the percentage of repairs made by various brand owners this could make a difference.

For example, if only 1,000 of those 49,000 respondents were Honda owners and 100 of them reported repairs within that 12 month period, that means that Honda would have a 10 percent repair rate. But if 15,000 of the respondents were Chevrolet owners and 1,200 of them reported repairs, that is an 8 percent repair rate. Yet this distribution was not provided so it’s difficult to ascertain exactly how those percentages compare within each sample size of brand owners.

d) What do you think of the journalist's reasoning, above? 

In another argument, the journalist compares the self-report survey used by Chevy with two other sources of reliability data, Consumer Reports and J.D. Power: 

...in a 2018 ranking of brand reliability, Consumer Reports ranked both Honda and Toyota well above Chevrolet, [which] occupied the lower part of the list along with several other domestic brands.

But like Chevrolet’s survey, this data is also self-reported based on the ownership experiences of Consumer Reports subscribers.

I also reached out to J.D. Power, as they are a respected reference when it comes to these kinds of “quality” rankings and asked them to comment on the methodology behind Chevrolet’s study. They declined. (Recall that Chevy has touted J.D. Power awards in this very same series of ads.)

e) Why might the journalist extol Consumer Reports or J.D. Power over Chevy's Ipsos survey? What might these organizations doing differently--if anything--in their measurement of reliability? 

f) Reflect on how the availability heuristic (the pop-up principle), cherry picking, bias blind spots, or the present-present bias (all from Chapter 2) might lead individuals to make faulty conclusions about which car brands are most reliable.

 

Thanks to Marianne Lloyd for sharing this idea! 

 

Claim: Mothers' friendships are good for babies' brains

The study showed that moms with larger social support networks had kids with higher cognitive test scores. Photo: Noriko Cooper / Alamy Stock Photo

The  Washington Post headline reads, "Mothers' friendships may be good for babies' brains."  The story is about an empirical study on mothers' social networks. Although it says "babies' brains", the study measured the babies' cognitive skill (there were no baby brain scans). The empirical article came out in JAMA Network Open. Does the original study support the claim? Let's find out.Before reading on, answer these questions:

 

a) What are the two variables in the Washington Post headline claim? What makes the claim a causal one? 

b) Of course, we need to conduct an experiment to definitively support a causal claim. Would an experiment be possible in this particular case?  

Here are some details about the study, as summarized by the journalist (not the researcher):

For the current study, researchers examined data on 1,082 mother-child pairs. They questioned women about their family structure, friendships and relationships in their communities and also looked at test results from cognitive assessments done when children were 2 years old.

Overall, mothers had an average of 3.5 friends in their social support networks. The kids of mothers with more than that tended to have higher cognitive test scores than the kids of those who had fewer, suggesting “network conditions were significantly associated with early cognitive development in children,” the study authors wrote.

c) Sketch a  simple scatterplot of the result described above. What will you put on the two axes? How many dots should your scatterplot have? What does one dot represent? 

Read the following details if you've studied Chapter 9: 

The researchers accounted for some factors that might influence child cognitive development, including mother’s age and IQ, father’s education level, and the child’s birth weight. 

In addition, "researchers also accounted for poverty levels"

d) The paragraph above is describing a multiple regression analysis. What would be the dependent (criterion variable) in this analysis? What are the predictor variables? (Hint: There are six predictor variables).

e) Imagine that someone proposes income as a possible internal validity problem. Given what you know about the study, can we say that "Perhaps friendship network size and child cognitive skills are correlated because of the third variable of income: That is, higher-income mothers may have more friends and more cognitively skilled children"?  Would this criticism be reasonable? Why or why not?

f) The journalist writes: "...they did not have data to assess other personal differences, such as history of depression, between mothers with large networks of friends and those with fewer."
This means history of depression could be a possible third variable explaining the relationship. How could the researchers rule out this third variable? 

The journalist writes: 

All of the families in the study lived in the area around Memphis. It is possible that results might be different elsewhere.

g) Which big validity is the journalist addressing in the statement above? 

h) Do you think the correlation between friendship and cognitive test scores would actually come out differently in a different city or country? Why or why not? 

 

Suggested answers:

a) The two variables are Amount of mothers' friendships and Level of health of babies' brains. The language of "may be good for" makes this a causal headline (remember we ignore adverbs such as "may" when classifying claims). 

b) Probably not; you'd have to randomly assign some mothers to have more friends and others to have fewer friends.

c)  One axis should have "number of friends" and the other should have "cognitive test score". There should be about 1082 dots on the scatterplot (better count your dots now...), and each dot represents a mother-child dyad.

d) The criterion variable is cognitive test score. The predictor variables are number of friends, poverty levels, mother’s age, mother's IQ, father’s education level and the child’s birth weight. 

e) Since the researchers did control for poverty levels, we can actually rule out wealth as a possible third variable. 

f) the researchers could obtain history of depression data on each mother and then add this variable as a predictor variable to the regression analysis. If the beta associated with "number of friends" is still significantly predicting cognitive test score, then we can rule out this possible confound (/third variable/ internal validity problem).

g) external validity

h) In my opinion, it seems unlikely that such a relationship is limited only to Memphis residents. 

Note: You can read the original empirical article in JAMA, here. Note to savvy instructors: The p-curve on the results in this paper suggests that there may be little evidentiary value. 

More butter, happier citizenry

What internal validity problems prevent us from claiming that butter leads to happiness? Photo: Fatty's place/Shutterstock

A writer for the Big Think wrote about how "Butter supply and life-satisfaction are linked".  It's a dairy interesting story with perfect scatterplots for practicing correlational designs.

Writer Frank Jacobs used the following bullet-points to open his story: 

• Haiti and other countries with low butter supply report low life satisfaction.
• The reverse is true for countries such as Germany, which score high in both categories.
• As the graph below shows, a curious pattern emerges across the globe. But is it causation or correlation?

You should click on the story and scroll down to the first scatterplot. 

a) What does one dot represent on this scatterplot?

b) What kind of correlation is being represented here: positive, negative, or zero?

c) Find your own home country. Is your country above the median or below the median on butter supply? Is your country above the median or below the median on life satisfaction? 

d) What would happen to the scatterplot if it were drawn with butter supply on the y-axis and happiness on the x-axis? Would it have the same shape and slope? 

The writer states:

Yet as the map shows, an abundance of butter does make people happy. Or could it be a case of correlation instead of causation? In that case, something else influences both life satisfaction and the availability of butter to go up and down together. 

d) In the curious paragraph above, the writer makes a clear causal claim. What is it?

e) After making the causal claim, the writer backs off, stating that "something else" could be associated with both life satisfaction and the availability of butter. He's talking about third variables here--or internal validity. What possible third variables might plausibly be correlated with both butter supply and life satisfaction in the world's countries? 

To help you consider possible third variables, zoom in on the countries that are in the lower left and upper right quadrants of the butter scatterplot. What dimension seems to separate the countries in those two quadrants? 

f) Take a look back at the column in Big Think. Is this data (and scatterplot result) from a published, peer reviewed study in an empirical journal? 

I butter thank Sue Frantz of Highline College for sharing this story with me. 

The science behind sleep

The science on sleep suggests that most of us need more of it!
Photo: Carlos Caetano/Shutterstock

A Washington Post article recently proposed that lack of sleep might be a "public health crisis." The journalist's story covered several types of empirical studies and made a series of frequency, association, and causal claims. 

Here are a series of quotes from the article. As you read each quote, ask yourself, a) How many variables seem to be involved in this claim? b) What is/are the variable(s)?  and then c) Is this claim frequency, association, or cause? How do you know? (Note: the links in each claim take you to the empirical articles on which they are based.)

Quote 1:

"Preschoolers who skip naps are worse at a memory game than those who snooze."

Quote 2:

"...a single night of sleep deprivation boosts brain levels of the proteins that form toxic clumps in Alzheimer’s patients." 

Quote 3:

"In 2016, the Centers for Disease Control and Prevention reported that a third of adults fail to get the recommended seven hours."

Quote 4:

"...even modest sleep reductions are linked to increased feelings of social isolation and loneliness."

Quote 5:

"A small study presented at the Society for Neuroscience conference last year found early evidence that playing “pink” noise — similar to white noise but with different frequencies — in a specific pattern could increase the proportion of time people spent in restorative deep sleep."

Quote 6:

"Adults over 50 with lots of insomnia symptoms were more likely to fall than those without, according to one study.:

Quote 7

"Studies have linked pain to poor sleep.."

 

Suggested answers to odd-numbered quotes.

Quote 1.

1. There are two variables:
2. Skipping a nap (or not) and level of skill at a memory game
3. This is an association claim, because the verb “are” is simply linking one type of preschooler (those who skip naps) with having lower skill at the memory game.

 

Quote 3.

1. There is one variable:
2. Getting 7 hours of sleep a night (or not)
3. This is a frequency claim. There is one variable and the sentence is mainly describing the number of U.S. adults who get that amount of sleep.

 

Quote 5.

1. There are two variables:
2. Playing “pink” noise (or not) and amount of time spent in restorative sleep
3. This is a causal claim because the verb used was “increase”.

 

Quote 7.

1. There are two variables:
2. Level of pain and quality of sleep.
3. The verb, “linked” is a verb signaling an association claim.

Which comes first? Cross-lag study on depression and social media use

Which comes first, social media use or depression? Photo:Africa Studio/Shutterstock

We have seen multiple studies that have tried to estimate how strong the link is between social media use and depression among young adults. We've covered some of these studies on this blog, including the effect size of the correlation and even a brain imaging study that manipulated "likes."

One challenge with studying real-world links between social media use and depression is that most use correlational designs. Therefore, we can't be sure whether any link between social media use and depression means that social media leads to more depression, or if depression leads to more social media use. And of course, in correlational studies there may even be a third variable, such as age, which is associated with both variables and causing them to appear correlated. 

The latest study used a longitudinal, cross-lag panel design, such as those described in Chapter 9, to study this question. Here's a writeup of the study by a journalist at Earth.com.

First, here are some comments about the design: 

The researchers point out that previous studies that found links between depression and social media use among tweens are limited in their results because they only captured a snapshot of teen social media use.

“You have to follow the same people over time in order to draw the conclusion that social media use predicts greater depressive symptoms,” said Taylor Heffer, the lead author of the study. “By using two large longitudinal samples, we were able to empirically test that assumption.”

Here are some more details about the method. They actually used two longitudinal samples: 

For this study, the researchers conducted a two-year longitudinal study surveying 6th, 7th, and 8th graders in Canada. The team also surveyed a series of undergraduate students once a year for six years as they progressed through college.

What were the results? 

Higher depression symptoms predicted social media use among teen girls which indicates that unhealthy social media use may be a symptom depression, not a catalyst for it.

“This finding contrasts with the idea that people who use a lot of social media become more depressed over time. Instead, adolescent girls who are feeling down may turn to social media to try and make themselves feel better,” said Heffer.

Questions

a) In a cross-lag panel design, researchers measure the same two variables in the same sample of people over time. What are the two variables they focused on here?

b) Sketch a series of boxes depicting the design of the study, similar to the figures in Chapter 9's cross-lag panel section. How many time periods would you need to depict for the undergraduate sample? 

c) Focus only on the girls. What do you think the autocorrelations might have been in this design? Fill those in on your sketch.

d) Still focusing only on the girls, given the description of the results, estimate what the two cross-lag correlations might have been. Which correlation should be larger--the one from Time 1 depression to Time 2 social media, or the one from Time 1 social media to Time 2 depression?

 

How much do chimpanzees cooperate?

Which behaviors were used to operationalize cooperation in chimps?
Photo credit: apple2499/Shutterstock

Human beings are a fundamentally prosocial and cooperative species, but what about one of our close relatives, the chimpanzee?  A science writer summarized a series of studies on chimpanzees who live in the Tchimpounga Chimpanzee Sanctuary, in the Republic of the Congo. The summary is here and you can read the empirical journal article here.

The studies illustrate how researchers can operationalize a variable like "cooperativeness" in other species. It provides a clear example of observational research, too. 

Researcher Alexandra Rosati conducted three observational tasks on the sanctuary, all meant to operationalize the construct cooperativeness. The first task took place with chimp volunteers who were alone in a dormitory area. One experimenter grabbed another one's stick and threw it out of reach into the dormitory. The second researcher made a show of trying to get the stick back through the bars of the dormitory. Would the chimp help her out by bringing her the stick? The researcher conducted 10 trials with each chimp. You can see the task in action at this link--watch Video 2. The journalist writes:

On average, the chimpanzees, who received no reward for cooperating, chose to hand the out-of-reach object to the researcher just over 50% of the time – but the rate of cooperation varied significantly among the individuals, with some helping in all 10 trials and others not at all. Chimps who chose to help more often also chose to help more quickly, supporting the conclusion that more cooperative individuals tend to make decisions faster than more selfish individuals do, the authors wrote.

Notice that there are two findings here. The first is the simple result of how many chimps helped (that's a frequency claim result).The second result is a correlational pattern. 

a.) What is the frequency claim you can make here? 

b.) Reread the description of the "stick" study (above) and decide what two variables are correlated.

c.) Placing these two variables on the x and y axis, sketch a scatterplot of the result that is described. (Compare your sketch to Figure 2a in the empirical journal article.) 

d.) Comment on the operationalization of the construct "Cooperation". Why might the researchers have chosen this operationalization? 

Here's a second way they studied chimp cooperation. This operationalization asks whether the chimps would punish individuals who don't follow rules:

When they had an opportunity to punish a thief who stole food from them, the chimps often chose to collapse the experimenter’s table [by pulling out the table's false leg] to prevent the thief from consuming the ill-gotten treats. 

e.) You can watch the "stealing" task here, under Video 3. What behavior (by the chimps) was used to operationalize "cooperativeness" here?  Why might this behavior be considered as an operationalization of Cooperation?

Finally, another task used in the study presented chimpanzees with two options for food. This task is viewable in Video 1 here: 

When presented with two sets of plates, one that contained a snack for both the chimpanzee and a familiar caretaker and another that had a treat only for the chimp, the apes were basically indifferent to whether the experimenter got something to eat. 

f.) The three operationalizations of cooperation in chimps found variable results. Two operationalizations showed some "cooperation," and one apparently did not. What does this pattern of results suggest to you about the importance of how well we operationalize variables? 

 

 

Reading vs. listening

Photo: Ladida/Getty Images

Audiobooks are more popular than ever. Have you ever wondered if there are cognitive differences between reading print books and listening to audiobooks? In this editorial, psychologist Dan Willingham explains some of the research comparing the two formats. Dr. Willingham describes several research findings, most of which come from experimental designs. 

Here's one point Willingham makes: 

...there are differences between print and audio, notably prosody. That’s the pitch, tempo and stress of spoken words. “What a great party” can be a sincere compliment or sarcastic put-down, but they look identical on the page. Although writing lacks symbols for prosody, experienced readers infer it as they go. In one experiment, subjects listened to a recording of someone’s voice who either spoke quickly or slowly. Next, everyone silently read the same text, purportedly written by the person whose voice they had just heard. Those hearing the quick talker read the text faster than those hearing the slow talker.

a. In the study above, what is the independent variable?  Was it manipulated independent groups or within groups (and what keywords in the description help you decide?)

b. What is the dependent variable? 

c. Sketch a small graph of the results described, labeling your axes mindfully.

 

Here is another study from Willingham's summary:

It sounds as if comprehension should be easier when listening than reading, but that’s not always true. For example, one study compared how well students learned about a scientific subject from a 22-minute podcast versus a printed article. Although students spent equivalent time with each format, on a written quiz two days later the readers scored 81 percent and the listeners 59 percent.

d. In this study, what is the independent variable?  Was it manipulated independent groups or within groups (and what aspects if the description helped you decide?)

e. What is the dependent variable? 

f. Sketch a small graph of the results described, labeling your axes mindfully.

g. Do you think you could have conducted this study as a repeated-measures design? Why or why not? 

h. What questions do you have about the internal validity of this study? 

 

Suggested answers

a. The IV is reading speed of the narrator during the first task; its levels are Slow and Fast. The phrase, "either spoke quickly or slowly" tells you this was manipulated as independent groups.

b. The dependent variable was how fast people read the text passage. 

d. Here the IV is reading a text or listening to a podcast. The results phrase refers to "readers" and "listeners," implying that the IV was manipulated as between groups (independent groups).

e. The DV is score on a comprehension quiz.

g. It might be difficult to conduct this exact study as a repeated-measures design, because people would have to process the same content aurally and via reading. Once they've processed the content once, it would undoubtedly be easier to process it a second time. 

h. You'd want to be sure that i) people were randomly assigned to the two conditions. This would control for selection threats; and ii) people in the two conditions read or listened to the exact same text (for example, a podcast and its transcript). This would help avoid a design confound.  

Can puppy pictures help your relationship?

Does pairing this cute puppy with a photo of your partner help 
improve your relationship?
Photo credit: Matanya/Wikimedia Commons

Professor James McNulty, a social psychologist who studies relationships, was asked by the Department of Defense to develop an intervention to help couples cope with the separation of a deployment. How can couples stay close when their partner is far away for months at a time? 

McNulty and his team of students tested an intervention in which couples viewed photos of their partner placed next to cute images of puppies, kittens, or baby bunnies.  Would just seeing your spouse next to cuteness make you feel closer to your spouse? Here's how a journalist from National Geographic described the study: 

For six weeks, one group of partners was shown a stream of images every three days that repeatedly paired their partner's image with that of a positive image, like a puppy. The control group saw pictures of their partners paired with neutral images, such as a shirt button.

Every two weeks [beginning at the start of the study], the couples' attitude toward each other were measured by the positive and negative words used when reviewing a series of faces, of which their partner was included. When couples were evaluated at the end of the study, those who had been shown positive images had more positive associations with their partners and improved marriage quality.

There were 144 couples in the study. You can read the empirical journal article here, in the journal Psychological Science. 

Questions.

a) McNulty and his team conducted an experiment. What are the independent and dependent variables? Was the independent variable manipulated as within groups or independent groups? 

b) Which type of experiment does it appear to be, based on the description: Posttest-only? Pretest-posttest? Concurrent measures? Repeated measures? 

c) Think about the design of this study and consider the 12 internal validity threats presented in Table 11.1. Which threats can you probably rule out, given the design of this study? Which threats do you think might still be in question? What questions might you ask about the study to assess these remaining threats to internal validity? 

 

Suggested answers:

a) The independent variable was which type of treatment they were in: Pairing their partner's face with cute animals or with neutral objects. This was an independent groups IV. The dependent variables included the positive and negative emotions people felt, and their marriage quality. That is, there were multiple dependent variables in the study.

b) This appears to be an independent groups design; specifically, a prettest-posttest design.

c) The researchers randomly assigned couples to the two groups, so we can rule out selection effects. It's not a repeated measures design, so we don't need to worry about order effects. Because they used a control group and an experimental group, we can rule out threats such as maturation, testing, regression, instrumentation, and history, because these threats would have applied equally to both experimental and control groups and thus evened out. We might ask about attrition (selection-attrition) in the two groups in case the unhappiest couples dropped out of the experimental group more often (however, if attrition was similar in the two groups then attrition is not an internal validity threat). We might wonder if some couples guessed the hypothesis of the study and changed their answers (experimental demand). Since both groups received a treatment (either the cute photos or the neutral photos), we can probably rule out a placebo effect. The dependent measures appear to be self-report, so observer bias is unlikely here. 

How well do fire alarms work for kids?

 Which fire alarm worked best to wake young children up? Photo: YMHhappyfamily/Shutterstock

An important applied psychology question concerns smoke alarms. Does the loud, screeching sound of a traditional smoke alarm work on kids? If not, what would work better? There's a Reuters press release about the findings of the study. 

Sleeping children, as it turns out, are fairly impervious to the screeching of a smoke alarm. Researchers found that most children ages 5 through 8 took more than five minutes to wake up with a standard alarm, as compared with around four seconds when they heard the sound of their mother’s voice, according to the results published in The Journal of Pediatrics.

“The thing that was most remarkable to us was to see a child sleep five minutes through a very loud high-pitched tone, but then sit bolt upright in bed when their mothers voice sounded through the alarm,” said the study’s lead author, Dr. Gary Smith.

In this research, experimenters invited families into a sleep laboratory on four different nights. Here's how a journalist summarized the study: 

Smith and his colleagues recruited 176 children between the ages of 5 and 12 and brought them into the sleep lab for testing. The lab had rooms that were set up to look like a typical bedroom, Smith said.

The children’s brain activity was monitored with electrodes on their scalp and face, and they were allowed to fall into the deepest stage of sleep. Then one of four types of simulated smoke alarm was sounded and researchers measured how long it took for a child to wake up and follow a previously rehearsed escape plan.

The experiment was run four times with each child, each about a week apart, so that all the children were exposed to all four types of alarm: standard high-pitched tone, the mother’s voice with instructions like “wake up” and “get out of bed,” the mother’s voice saying the child’s name, and the mother’s voice saying the child’s name and then giving instructions.

There was no significant difference in time to waking between the three alarms that used the mother’s voice. But there was a very large difference between all three of the maternal voice alarms and the standard alarm.

 

Questions: 

a) This is a simple experiment. What is the independent variable? What are its levels? Is it manipulated as independent groups or within groups? 

b) What is the dependent variable? 

c) Which of the four simple experiments is this: Posttest only? Prettest-posttest? Repeated measures? Concurrent measures? 

d) Sketch a graph of the results of this study.

e) What are the advantages to using a within-groups design in this study? 

f) Which internal validity threat do we most worry about in a design like this? How should the researchers control for that internal validity threat?

 

 

Suggested answers:

a) The independent variable is which smoke alarm they tried, and it has four levels: Traditional beep, Mom's voice saying the child's name; Mom's voice saying the name and giving instructions, and Mom's voice giving instructions. It was manipulated within groups. 

b) The DV is time to wake up (and whether they followed the right instructions)

c) This is a repeated measures design. 

d) Sketch should have "time to wake up" on the y axis and the four alarm conditions on the x-axis. 

e) The strongest advantage here is that each child can serve as their own control. Each child's normal level of "wakeability" is held constant across the four alarm conditions. 

f) We mainly worry about order effects in most repeated measures designs. The researchers should counterbalance the order of the four conditions to help control for this threat. To find out if they did that, refer to Method section of the empirical journal article, here. 

"Join a team and get sweaty" to help with depression

Photo credit: isitsharp/Getty Images

A health story's headline reads, "A New Prescription For Depression: Join A Team And Get Sweaty"

Psychologists have known for a while that depression levels are linked to exercise. In fact, some comprehensive treatments for depression include an exercise component in addition to psychotherapy or medication. 

This recent NPR story focused on which type of exercise that might work best: 

..a recent study in the journal Lancet Psychiatry found that popular team sports may have a slight edge over the other forms of physical activity.

The researchers analyzed Centers for Disease Control and Prevention survey data from 1.2 million adults and found — across age, gender, education status and income — people who exercised reported fewer days of bad mental health than those who didn't. And those who played team sports reported the fewest.

a. What are the two central variables in this research, based on what you've read so far?

b. Is this study correlational or experimental? 

c. Sketch a graph of the result. 

d) The journalist writes: 

Now, this study only shows an association between group exercise and improved mental health, and can't prove that the one causes the other. 

Why can't the study support that "one causes the other"?  Apply the three causal criteria (covariance, temporal precedence, and internal validity) to this study.