162 CHAPTER 5
a person was when a loss occurred, the more symptoms of depression he or she is
likely to have after a breakup as an adult. However, this relationship may not hold
for every participant.
You could determine whether the correlation coeffi cient is statistically signifi cant
through a calculation that depends partly on the number of data points you have. All else
being equal, the more data points, the smaller the correlation coeffi cient needs to be
to count as statistically signifi cant. Why is this so? Consider the fi nding for gifted
children of negative correlations between measures of self-esteem and depression and
also between measures of self-esteem and aggression (in other words, gifted children
who had lower self-esteem also tended to be depressed and to be more aggressive;
Benony et al., 2007). If this study had involved only 4 gifted children (instead of
the 23 actually examined), the correlations could easily refl ect the luck of the draw:
Perhaps the day before the study one of the 4 children had misread the instructions
for an important test, which he then failed dramatically. That experience might have
produced temporary low self-esteem and also made him feel depressed and aggres-
sive. A chance event would have made the sample for this study nonrandom. Alter-
natively, suppose that one of the 4 children had absolutely aced an important test at
school the day before—and so was feeling especially good about herself and not de-
pressed or aggressive. This too could distort the results. In contrast, if the study had
100 children as participants, the chances that some would have had an experience
that produced temporary low self-esteem would likely be balanced by the chances
that some would have had an experience that produced the opposite effect. The bot-
tom line: The larger the sample, the more likely that random variations going in one
direction will be canceled out by random variations going in the other direction.
Thus, the larger the sample, the smaller the correlation coeffi cient needs to be for
you to be confi dent that the observed relationship is not a result of chance.
Researchers want to know not only the correlation coeffi cient, but also the
value of p (which stands for probability) that is associated with that coeffi cient;
the value of pindicates how likely it is that the correlation could have arisen due to
chance. In the study just mentioned, the correlation between a measure of self-esteem
and a measure of depression was r = –.59, and this coeffi cient was tied to p < .01. This
means that the probability that the correlation is due to chance is less than 1 in 100.
Similarly, a value of p < .05 means that the probability that the correlation is due
to chance is less than 5 in 100. In fact, p < .05 is usually considered the cutoff for
statistical signifi cance. You can check the signifi cance of the value of p for a data set
by consulting a published table (which appears in appendixes of most statistics text-
books). Computer software that calculates the correlation coeffi cient usually gives
the value of p as well.Using Correlational Methods
Much of the research on defi ning and understanding psychopathology is correla-
tional. That is, it is designed to discover whether one variable (a disorder or a symp-
tom) is linked to other variables (such as alterations in neural activity, irrational
thoughts, or family functioning). Epidemiology is a type of correlational research
that investigates the rate of occurrence, the possible causes and risk factors,
and the course of diseases or disorders. Thus, in epidemiological studies of
psychopathology, researchers identify people with one (or more) disorder
and correlate the presence or severity of the disorder with other variables,
such as the age of onset of the disorder, the number of people in the family
who have had symptoms of the disorder, or socioeconomic status.
Note that certain factors (such as having a relative with a disorder) can be
risk factors, which increase the likelihood of developing a disorder. However,
by defi nition, risk factors are simply that—risks, not destiny. How are such
risks identifi ed? Studies often use correlational data to determine whether peo-
ple who have a psychological disorder are different in some way from people
who don’t have the disorder. Some of these studies are longitudinal studies,
which are designed to determine whether a given variable is a risk factor byResults of epidemiological research indicate that
people who engage in pathological gambling—
their gambling is compulsive—are at high risk to
have an alcohol or drug problem (Petry, Stinson,
& Grant, 2005). However, fi nding such a correla-
tion does not tell us why there is a relationship
between gambling and substance use problems.Getty Images
Epidemiology
The type of correlational research that
investigates the rate of occurrence, the
possible causes and risk factors, and the
course of diseases or disorders.Longitudinal studies (in studies of
psychopathology)
Research studies that are designed to
determine whether a given variable is a risk
factor by using data collected from the same
participants at various points in time.