Topic: Consider these statements:
“Qualitative research methods are the best!”
“Quantitative research methods are the best!”
Thread Prompt: Present an argument in your thread that supports EACH of these 2 statements. Separate your writings (do not mix these 2 statements as 1 statement), and label them as “Statement 1” and “Statement 2” in your thread. Support each of your statements with content from your textbook. There is more than 1 way to answer this correctly; use imagination and creativity, and support it with content from the textbook.
Introduction to Research: Less Fright, More Insight: A Customized Version of Research Methods: Are You Equipped? Second Edition by Jennifer Bonds-Raacke and John Raacke. Designed specifically for Ray.***
OUTLINE Are You Equipped? Differentiating among Methods Non-experimental Methods Factors to Consider Converging Research Methods Are You Equipped Now? Chapter Summary APA Learning Goals Linkage ARE YOU EQUIPPED? How do you pick what restaurants to eat at when you are traveling? Many people we know visit a website called Yelp (www.yelp.com). At this website, people can literally rate restaurants by providing a review with comments and the website will also filter restaurants by features such as distance from your location, price, and neighborhood, to name a few. You can read the reviews and use the filters to select a restaurant to visit, especially when you are in an unfamiliar city and want to reduce the risk of picking a bad place. The use of this website (and others which are simi-lar) brings up an interesting question. Specifically, is the price point of an establishment related to customer satisfaction? In other words, does what you pay for a meal impact your satisfaction with the restaurant? How could we answer this ques-tion? Take a minute and think about how you could gather information to address this question. What would you do? There are numerous ways to gather information to find out if price of a restaurant is related to cus-tomer satisfaction with the meal. The fact that there is more than one way to gather information on this topic illustrates that social scientists have a plethora of research methods at their disposal. In this chapter, we are going to focus on non-experimental research methods. We could use non-experimental research methods to examine the relation-ship between price and customer satisfaction in several ways. First, we could use the existing records provided by the websites mentioned above. This would allow us to see if the two variables were related to one another. Using exist-ing records such as those from yelp.com is known as archival research. Another way to see if the two variables were related to one another would be to ask customers. You could have customers complete a questionnaire asking about their opinions as they left different restaurants. This is known as survey research. Archival and survey research are just two examples of non-experimental research which seeks to examine relation-ships among variables. We will talk in more detail about these methods and others throughout the chapter. Before we begin, you might want to know that research has been conducted on this topic and a relationship does exist between price and customer satisfaction. For example, Ye, Li, Wang, and Law (2014) found price was related to cus-tomer satisfaction when examining reviews of hotels through online reviews. As you go through this chapter, we also want you to keep in mind how the material relates to the American Psychological Association goals for psychology majors. Specifi-cally, this chapter will address the following goals: • APA Goal. Scientific Inquiry and Critical Thinking You will demonstrate scientific reasoning and problem solving, including effec-tive research methods. • APA Goal. Ethical and Social Responsibility in a Diverse World Communication You will demonstrate competence in writing and in oral and interpersonal com-munication skills. DIFFERENTIATING AMONG METHODS Qualitative: Research methodologies where researchers seek insight through in-depth collection of information, hopefully resulting in new hypotheses.* Quantitative: Research methodologies that seek objectivity through testable hypotheses and carefully designed studies. Psychology is the scientific study of thoughts and behaviors. Therefore, psychologists, and social scientists in general, use a variety of research methods to study what people think and do. There are two main ways to differentiate among types of research. First, you can describe a research method as being either qualitative or quantitative. Quali-tative research uses inductive logic in the various qualitative approaches. Researchers using qualitative methods seek insight through in-depth collection of information that maybe subjective, hopefully with hypotheses emerging from the in-depth information.* Researchers gather the data and draw conclusions on the basis of their observations. On the other hand, quantitative research is deductive by nature. Researchers using this methodology seek objectivity through testable hypotheses and carefully designed stud-ies. Researchers gather data that can be reported in numbers and statistics. Conclusions are drawn from statistics and generalized to populations of interest. It is interesting that researchers can be trained in either qualitative approaches, or quantitative approaches, or both. Dr. Crawford, for example, had considerable train-ing in both approaches during his Ph.D. program. The two approaches need not be viewed in opposition to each other, and both can shed light on understanding variables of interest. Quantitative approaches will want to quantify what is happening with the variables in the study, whereas qualitative approaches can help discover, identify and more richly describe what the variables really are that need to be studied! Another way to differentiate among types of research is by describing the research as experimental or non-experimental. Experimental research involves the manipulation of a variable or variables being studied, and assigning participants to treatment con-ditions. We will spend more time in later chapters talking about how to design and interpret experimental research, and how to identify limitations of the results. For now, know that variables are identified, one or more variables are going to act on the depen-dent variable to see if something will change in the dependent variable. The design of the experiment (also called experimental design) is crucial in determining if the results of the study are reliable or valid. More on those topics later in the book. Non-experimental research does not rely on manipulation of variables. Rather it makes observations about how variables are related to one another, describing those findings. And, it can even make observations that will help discover what new variables are, new variables that can be studied in various types of research
.* INTRODUCTION TO CORRELATIONAL METHODS
As mentioned above, non-experimental research does not manipulate variables of interest. However, even without direct manipulation, you can still explore relationships between variables using correlational research methods. Correlational research meth-ods are very important to the field. This methodology provides us with information on the initial link between variables of interest. Correlational research methods are also frequently reported in the media. We want to tell you about two correlational studies recently reported in the media examining issues related to weight and obesity. The first study was conducted by Kopycka-Kedzierawski, Auinger, Billings, and Weitzman (2008) at the Eastman Dental Center, which is part of the University of Rochester Medical Center. Kopycka-Kedzierawski et al. found that for children between the ages of 6 and 18 there was a rela-tionship between weight and tooth decay. Surprisingly, as weight increased, the risk of tooth decay decreased. We will talk a little later on about why these results might have been obtained. For now, we want to focus your attention on how this study shows that two variables (i.e., weight and tooth decay) are related to one another. Our second news story reported that Gazdzinski, Kornak, Weiner, Meyer-hoff, and Nat (2008) found that weight was related to bio-chemical deficiencies. Specifically as weight increased for individuals, researchers found lower-than-normal levels of markers for neuronal health and cell membrane syn-thesis. We will return to understanding these results in just a minute when we discuss the advantages and disad-vantages of this method.
When you have a correlational research method, your results will tell you if the two variables are related. Variables can be either positively or negatively related. Do not be confused by the labels of positive correlation and negative correlation. This does not mean that positive correlations are good and negative correlations are bad. Rather, positive correlations have variables that vary in the same direction and negative cor-relations have variables that vary in opposite directions. To make this clear, we want to tell you about some examples of positive and negative correlations. We have also provided you with the figure below to visually represent each type of correlation. Two variables that are positively correlated are years of education and salary. This is a pos-itive correlation because as scores on one variable increase so too do scores on the second variable. Similarly, as scores on one variable decrease so too do scores on the second variable. In this situation, the more number of years of education a person has, the higher the salary. Conversely, the fewer number of years of education a person has, the lower the salary. In both instances, the two variables either increased or decreased together, making this a positive correlation. Another example of a positive correlation is time spent studying for a test and performance on the test. In general, the more you study for a test, the higher your grade on the test, and the less you study for a test, the lower your grade on the test. Again, this is a positive correlation because both variables are increasing or decreasing together. Another example of a positive correlation is the amount of food consumed and weight. The more food that you consume, the higher your weight and the less food that you consume, the lower your weight.
With negative correlations, the two variables of interest are related to one another as well. However, as one variable increases, the other variable decreases, or as one variable decreases, the other variable increases. An example of a negative correlation is marital satisfaction and likelihood of divorce. As marital satisfaction increases, the likelihood of divorce decreases, and as marital satisfaction decreases, the likelihood of divorce increases. The two examples that we presented earlier from news reports were also examples of negative correlations. Remember that as weight of children increased, the risk for tooth decay decreased. The reverse can also be stated. As the risk of tooth decay increased, the weight of children decreased. Thus, these two variables were related to one another but in opposite direc-tions. Similarly, in the second report, as weight of individu-als increased, normal levels of brain functioning decreased, or as normal functioning increased, weight of individuals decreased. We find it helpful to draw arrows for each of the variables that we are reading about. If the variable is increas-ing, we draw an up arrow. If the variable is decreasing, we draw a down arrow (Figure 4.1). We can then look at the arrows to see if they are both pointing in the same direc-tion (positive correlation) or in diff erent directions (negative correlation).
3. This is also an example of a positive correla-tion because both variables (sense of unfairness and ratings of pessimistic view of the future) are increasing together. Also, note the reverse state-ment is true: as sense of unfairness decreases, pessimistic views of the future decrease. 4. Finally, there is a negative relationship between sense of unfairness in the childhood home and self-esteem. As one variable increases (sense of unfairness), the other variable decreases (self-esteem). This brings us to an interesting point to consider. Specifically, how do you interpret these correla-tional research findings? Should you hold your baby all the time so he or she does not cry? Not necessarily. We will dis-cuss three questions for you to ask yourself when interpreting correlational research findings. How did you do with these four examples? We should mention that sometimes the direction of a variable is not as clearly stated in the media or in a research article as we did in our examples. However, if you stop and think about the results, you will be able to determine if the scores are increasing or decreasing for the variable.
As with any research method, there are advantages and disadvantages to examining correlations between variables. One major advantage of correlational research is that it allows us to make predictions. For example, if we know marital satisfaction and likeli-hood of divorce are negatively correlated, it can help us in counseling couples who are experiencing low marital satisfaction. Many times, examining correlations between variables is a great starting point to researching a topic. Plus, it is a useful method when conducting a true experiment would not be ethical. This brings us to a limitation of the method, determining cause and eff ect. Using a previous example, should we stop buying ice-cream so that we can reduce the number of murders committed? This even sounds like a strange question to ask. When you have a correlation, you must think about the directionality of the correlation and ask yourself the following questions: • Is X causing Y? • Is Y causing X? • Is there a third variable causing both X and Y to be related? In the example of ice-cream sales and murder rates, you would ask yourself the follow-ing questions: • Does eating ice-cream (X) cause you to commit murder (Y)? • Does committing murder (Y) cause you to eat ice-cream (X)? • Is there a third variable that is causing ice-cream sales (X) and murder rates (Y) to be related?
SECTION SUMMARY •
Qualitative research methodologies seek subjectivity through in-depth collection of information and emerging hypotheses. • Quantitative research methodologies seek objectivity through testable hypoth-eses and carefully designed studies. • Experimental research is a class of research methodologies that involve direct manipulation of a variable. • Non-experimental research is a class of research methodologies that do not rely on manipulating variables. • Correlational research methods evaluate the relationship between variables. • Positive correlations have variables that vary in the same direction. • Negative correlations have variables that vary in opposite directions.
NON-EXPERIMENTAL METHODS So far, we have provided you with introductory information that will be helpful as you learn about non-experimental designs. In this chapter, we will discuss the follow-ing non-experimental designs: ethnography, naturalistic observation, case studies, archival research, content analysis, and survey. It is important to note that all these non-experimental designs can examine relationships among variables as mentioned previously under correlational methodologies. We will begin with ethnography. Typically, ethnographies are associated with the field of anthropology and are used to describe new cultures. The purpose of this method is to describe a culture in detail. In doing so, the researcher records and transcribes events that he or she witnesses and shares these findings with others. Many times, ethnographic questions concern the relationship between culture and behavior, and thus other social scientists are interested in the method. For example, Russell (2011) used ethnographic techniques to learn more about homeless women in Baltimore. One advantage of this methodology is that you can get “rich” or detailed information from an insider’s perspective. Disadvantages include the lack of a testable hypothesis, the inability to infer cause and effect, and little ability to generalize the results to other groups (to be discussed further in Chapter 5). Ethnography: Used to describe a culture in detail by recoding and transcribing events that are witnessed. Non-experimental
Naturalistic observation is where you observe people or animals in their natural set-tings. These observations can occur in the field (sometimes, called field studies) or in the laboratory (referred to as laboratory observations). Studying people in their natu-ral setting means many things. One of the most famous naturalistic observations was conducted by Jane Goodall. In the summer of 1960, Goodall went to East Africa to live among the chimpanzee population. There were many aspects about chimpanzees that Goodall wanted to learn such as if chimps used tools. She believed the best way to understand the chimp behavior was to observe them in their natural environment. Goodall’s work has led to numerous publications on the life of chimps. We want to follow up this classic example of naturalistic observation with more every-day examples to illustrate that natural settings are diverse and you are not required to travel to far-off destinations. First, Middlemist, Knowles, and Matter (1976) wanted to know how the presence of another man in the bathroom influenced men’s selection of a urinal and urinating behaviors. To do so, the researcher hung out in front of the bathroom mirror. The researcher gathered information on which urinals men used and their urinating behaviors (i.e., length of time to begin urinating and duration of urination). Results indicated that men prefer not to use a urinal next to another man. In addition, the closer a man is to another man in terms of urinal distance, the longer
it takes for them to begin urinating and the shorter the duration of urination. It would be interesting to see if gender differences occur in this behavior. Another example of a naturalistic observation comes from developmental psychologists who routinely use laboratory observations to study children. Important developmental information has been gained by bringing children into the lab and observing their interactions with their mom and dad. Examples include informa-tion on attachment style and stages of development such as object permanence. Finally, we use naturalistic observations in the world of teaching. It is not uncommon for fellow professors to observe one another in the classroom. These observations are used early on in a faculty member’s teaching career to provide them with constructive criticism and later on in a career to support decisions regarding promotion and tenure.
The three examples of naturalistic observation that we have shared with you are varied in topic and scope. However, there are some commonalities in these observations. To begin, the researcher had to decide on a topic of behavior to observe and whether his or her presence would be known or hidden to the participants. In Good-all’s situation, it was almost unavoidable that the chimps were aware of her presence. Similarly, when a professor has their teach-ing observed by a colleague, the professor and the students in the class are aware of the observation. In fact, students will quickly recognize a new face in the crowd. However, it was possible for the researcher in the urinal research example to hide his true purpose and pretend to be using the restroom too. Developmental psychol-ogists often hide their presence by videotaping childrens’ inter-actions with parents and responses to new stimuli from another room. Another commonality in our examples was that the researcher engaged in a systematic observation of specified activities. Sometimes, a coding system is used to describe the observed behaviors and researchers must be trained prior to their observa-tions. Finally, to carry out their observations, researchers will need equipment to note their observations and may use a video camera. There are many advantages to this methodology. To begin, the behavior being observed is natural and spontaneous. For the most part, participants being observed in natural-istic observations are just doing what they normally do in life. This is an advantage over research procedures that require people to participate in a lab setting or participate in tasks that are unfamiliar to them. It would be very difficult to answer questions about male bathroom behavior in a more realistic way. However, this method also has sev-eral disadvantages that you must consider. One main disadvantage is how the observer changes people’s behavior by his or her presence. For instance, we can talk about this limitation with the example of teaching observations. Every time a peer has observed our classes, the students act differently. They become shy and well-behaved compared to a typical day. This could be due to the fact that the students are trying to behave and make us look good to the observer or we are acting differently and the students are picking up on our changed behavior. Either way, the observation itself changed the natural setting. If a researcher decided to avoid this disadvantage by hiding his or her presence, he or she must consider the participant’s right to privacy as well. Another disadvantage is that the researcher has to wait for events to occur. It might have taken a few hours before a man came to use the restroom; Goodall waited months before seeing chimps use tools. Researchers also need to be careful not to introduce bias in their observations. For example, researchers might be looking for a particular behav-ior and report “seeing it” when others would not. A well-known example of this has occurred with observation research of chimps using sign language. Some researchers reported seeing animals use sign language, while other researchers reported the ani-mals were not signing. Finally, cause and effect cannot be determined from naturalistic observations.
YOU TRY IT! We have just covered information on naturalistic observations. If you wanted to examine how children play with one another and how frequently aggressive behaviors occurred, how could you use a naturalistic observation to investigate these topics? Discuss in your answer how you could conduct the study when your presence was known and unknown to the children.
ANSWER You could investigate how children play with one another and their aggressive acts during play by going to a place where children naturally play together, like a playground. This could be at a community park or at a school. You could hide your presence if you are worried that children will play differently knowing you are watching. For example, you could watch children on a school playground from a classroom window or you could take your own child to the park and be just another parent. On the other hand, you might decide that children knowing you are around will not terribly influence their behavior so you could just sit and watch. In making your observations, you would probably have devised a coding system in advance. This way you would know what was considered violent behav-ior. If possible, another person could watch with you and make observations, allowing for you to compare observations. Case Study: A research methodology that is an in-depth observation of an individual, animal, event, or treatment method. The next non-experimental research method that we want to discuss is a case study. A case study is an in-depth observation of an individual, animal, event, or treatment method. Typically, an intensive observation is done and detailed account is taken because the event is extremely rare and unusual. There are classic examples of case studies in the field of Psychology. These include Phineas Gage and cases of feral chil-dren. Phineas Gage provided us with information on the link between personality and parts of the brain. Specifically, in 1848, an explosion sent a tamping iron through Gage’s skull. Surprisingly, Gage survived the explosion but his behavior changed greatly due to the damage in the frontal lobes of his brain. Researchers have long been interested in providing a detailed account of what took place and how Gage’s behavior changed as a result. Cases of feral children have provided the field with information on the impor-tance of early exposure to language. For example, a child who is referred to as Genie was found around the age of 13. Until she was found, Genie had been kept in isolation and deprived of any exposure to language. This case study was an opportunity to see how language deprivation would influence Genie’s ability to later acquire a language. Genie was able to learn some level of English. However, her fluency was impaired and the actual level of her fluency is debated among experts (Curtiss, 1977). When telling students about case studies, we always mention the two classics above. In addition to these, we have two more case studies that we heard about during our studies and have not forgotten. The first case study is from the area of human sexuality (Linnau & Mann, 2003). A male patient was admitted to the hospital for severe abdominal pain. When questioned, the patient admitted to swallowing Barbie doll heads (including hair) for sexual gratification. Researchers were interested in understanding how this provided the patient with sexual gratification and how such a behavior develops. A final case study is that of Clive Wearing. Clive Wearing was a brilliant musician. However, due to a case of viral encephalitis, Clive lost the ability to form new memories. This is known as anterograde amnesia. The Mind, a series by BBC, documents the events leading up to and after Clive’s memory loss. Researchers are particularly interested in the parts of the brain damaged and the result on Clive’s memories and behaviors.
The major advantage to case studies is that we can study rare events that would be unethical to study otherwise. It would be extremely unethical to deprive a child of language, make people swallow objects, or damage parts of the brain to see how the event influ-enced an individual’s behavior. Therefore, case studies provide us with unique opportunities to better understand situations that we could not study experimentally. Despite this advantage, there are limitations to this method. First, we do not always know the cause of the behavior. In trying to understand the male patient admit-ted to the hospital, researchers attempted to reconstruct events from his past to explain his current situation. Yet, this procedure introduces bias because the researchers are selecting what informa-tion from his past they think is important. There could be other variables that the patient and the researchers were unaware of that influenced him. Second, these unusual events might not influence everybody in the same manner. For example, Gage was extremely fortunate to have survived the explosion. Not all people would have survived. Further-more, not all people would have experienced the same resulting behavior. Thus, when using case studies, it is important to keep in mind the limitations of the findings. Case studies to not always need to be just for the unusual or the extremely rare situa-tions. Case studies can be used for many kinds of practical applications. Suppose you are a therapist, and you are writing a report for your client’s lawyer to present to a judge; that is not necessarily an extremely rare or unusual situation. This is one more example of how the case study method of research can be used. Your report for the judge is essentially a specialized case study report about your client.* Archival research is a non-experimental method where you use existing records and select portions of the records to examine. These existing records were collected by other people as a form of public records. Some examples include census information, marriage applica-tions, police arrests, and reports prepared by your university. One of Jenn’s first studies as an undergraduate student was an archival study. Jenn was examining sex and age differences in couples applying for marriage. To do so, she obtained the local newspaper’s listing of couples applying for marriage licenses in Shelby County, Tennessee, for 1 month. This provided 783 couples. For each listing, the newspaper provided the name and age of the applicants with the male’s name listed first. Jenn found that in 63% of the couples the male was older than the female. She discussed these findings in relation to sociobiological theories (Bonds & Nicks, 1999). Jenn has continued to use archival studies in her career. For example, she has used existing records to (a) compare the number of women to men journal editors in psychology, (b) determine the level of engagement in the classroom of Native American students com-pared to other ethnic groups at an East Coast university, and (c) examine rates of graduation for honors students compared to nonhonors students. Within the field, archival research is used to investigate a variety of topics. For example, Granhag, Ask, Rebelius, Öhman, and Giolla (2013) used descriptions reported to the police by witnesses of a murder to look at accuracy of basic and detailed attributes of offenders and Brenner, et al. (2013) used an archi-val method to assess rates of brain injury for Veterans seeking psychological help.
SECTION SUMMARY • A variable is an event or characteristic with at least two attached values. • When conducting a study, there are two important variables to be considered: • The independent variable is the variable in a study manipulated by the researcher. • The dependent variable is the variable within a study observed or measured. • A subject variable is a characteristic or attribute of a participant that can impact the participant’s behavior or thought within a study. • Subject variables are often traits specific to a participant, such as sex, age, or ethnicity. • Treatment conditions refer to levels or number of groups in the independent variable. • The experimental group is the group exposed to the independent variable. • The control group is the group not exposed to the independent variable. • The placebo control group is exposed to an inert substance or object similar to the independent variable but having no effect
RELIABILITY AND VALIDITY In the first part of this chapter, we introduced you to the concept of variables and dis-cussed independent and dependent variables. When you read about independent and dependent variables, it is important to consider two measurement concepts. The first concept is reliability. Reliability deals with the consistency when measuring variables in a research study. Consistency is the important feature here. For example, the speed-ometer on Grampa Ray’s motorcycle is inaccurate as far as actual speed is concerned, but it is always off by the same amount. So, since it is consistent (although not perfectly accurate) it is still a reliable measure of speed.* The second concept is validity. Validity is concerned with the accuracy of the measurements used to assess different variables. Consequently, Grampa Ray’s motorcycle’s speedometer is not valid, because it does not accurately measure what it is supposed to measure, speed. It is still reliable, because it is consistent even though it is consistently inaccurate. But it is not valid, because it does not give an accurate measurement of speed.* In the remainder of this chapter, we will present you with specific techniques used to establish the reliability and validity of variables. We will begin with reliability.
RELIABILITY Reliability of a variable is very important when conducting research. Reliability is the consistency of your measure to produce similar results on different occasions. Therefore, reliability is primarily concerned with being able to replicate or reproduce the findings. To make sure a measure is consistent in its ability to evaluate a variable, there are several types of reliability assessments. Let’s look at some types of reliability that will be useful for evaluating the reliability of an assessment tool, for example something like the multiple choice tests you have for your courses, or other assess-ment tools.* Test–Retest Reliability: A reliability assessment where your measure is tested on two different occasions for consistency. The most common type of reliability assessment is known as test–retest reliability. When using a test–retest assessment, you give your measure to a sample of participants individually (test) and then again at a later date (retest). Usually, the testing is done a few weeks apart. In order to make sure the measure is reliable, you look at the two scores for each participant. If the measure is reliable, the individual will have compa-rable scores on the two points in time. For example, if we were to give an intelligence test to your class at the beginning of the semester and then again midway through the semester, we would expect to see similar scores for each person. We would not expect the scores to be exactly the same, just close. To determine whether the measure is reli-able, you compute a correlation coefficient for the scores. A correlation coefficient of 0.80 in a measure is generally seen as very reliable. At this point, you do not need to know how to compute a correlation coefficient. Rather, the goal is to have you prepared to know what to look for when reading about this procedure in research articles. One disadvantage of the test–retest procedure is that it is time consuming.
Another method which does not require the same amount of time to assess reli-ability is known as the split-half method. The split-half method occurs when you administer a measure to a sample of participants. Unlike a test–retest method, where you would wait several weeks before giving the measure again, the split-half method uses only the results from the first collection of data. Specifically, you would split the measure in half. It might sound strange to split a measure into half. However, consider assessing the reliability of a 50-question multiple-choice exam. This could be done by randomly assigning questions from the exam to two groups, dividing the exam between odd and even questions, or by dividing the exam at the midpoint. You then compute scores for each half finding a correlation coefficient between the two halves. One concern is the method in which you split the halves can impact the correlation. Specifically, if you unknowingly have several similar questions in the same half and none in the other half, the correlation will be low. Therefore, many researchers use a modified version of the split-half method known as internal consistency. Internal Consistency: A reliability assessment similar to the split-half method. However, the splitting occurs more than once and an average of the correlations is taken. The internal consistency method is the same as the split-half method with one excep-tion. In the internal consistency method, you repeat the split-half procedure multiple times, thus collecting multiple correlation coefficients. Then, you average the multiple correlation coefficients. This method counteracts the impact of having too many simi-lar questions in only one-half of the split. A commonly used statistic for computing internal consistency is Cronbach’s alpha. You will learn more about statistical tests in the statistics course you complete as part of your degree.*
Another way to assess the reliability of a measure is to use the parallel-forms method. In this method of reliability assessment, you divide a measure into two parts. Ques-tions from the original measure are randomly divided among the two parts. You then administer both of the parts to a sample of participants. Following the administra-tion, you compute the correlation coefficient for the two parts. The higher the cor-relation between the parts, the more reliable the measure is said to be. The benefit of this approach is the brief amount of time it takes to assess reliability. However, one disadvantage is you must generate a large number of questions in order to divide into two parts. The last assessment of reliability we want to discuss is interrater or interobserver reliability. The previous assessments of reliability were focused on how you would construct a measure (such as constructing an exam in a class). However, inter-rater or interobserver reliability is focused on using a measure consistently in research. Interrater reliability is used when a research design calls for observa-tions of an event. This type of reliability assessment is used to ensure the observa-tions being made are consistent and not biased. For example, if you want to evaluate bullying behavior in children, you could observe 100 instances of bullying and cat-egorize the bullying into one of the four categories. If you were the only researcher making the observations, there is no way to know if you consistently categorized the 100 observations. However, if there is more than one observer, you can become more confident in the ratings. Interobserver reliability uses more than one observer and the observations of the observers are compared to assess the level of agreement. In the bullying example, two or more observers would compare the categorizations of the 100 instances of bullying. If the observers were to agree 92 times out of a 100, the reliability would be 92%, which is quite reliable. However, if the observers only agreed 48 times out of a 100, the reliability would be 48% and the measure is not very reliable. Essentially, you are calculating a correlation coefficient between the degree of simi-larities in the observations of the observers. To be confident in the reliability of your results, you want to obtain a high degree of similarity. The most important consideration when using multiple observers, or raters, is that the observers themselves are consistent with each other. The raters will need appropriate training to know how to observe, what to look for, how to rate what is observed, etc. So to have good inter-rater reliability when you have more than one observer or rater, these observers need to have consistency in how they observe or rate. This takes train-ing, which also adds time and cost to your study.* VALIDITY As we have mentioned, selecting your variables is an important part to any research design, as is measuring them. In this section, we will talk about the concept of validity. Validity is defined as the ability of your measurement to accurately measure what it is supposed to measure. This is different from reliability, which is about being able to replicate scores on future instances. Let’s begin with an example
INTERNAL, EXTERNAL, AND CONSTRUCT VALIDITY Internal Validity: It is confidence in saying the observed change in the dependent variable is due to the independent variable and not due to any outside influences. When measuring a variable, there are several types of validity to consider. The three most common types of validity are internal validity, external validity, and construct validity. Each of these types of validity deals with a different aspect of measurement. However, the commonality in all is that they are concerned with the accuracy of the measures used in a research design. The first type of validity we will discuss is internal validity. Internal validity is an important factor for independent variables. Internal validity is confidence in saying the observed change in the dependent variable is due to the independent variable and not due to any outside influences. This allows you to make a causal inference regarding the influence of the independent variable on the dependent variable. This is important when conducting a study because you want to be able to show the manipulation had an impact. Rothbaum, Anderson, Hodges, Price, and Smith (2002) examined different types of therapies to relieve fear of flying. Specifically, over a 6-week period, participants were placed into one of the three groups, where each group was exposed to a different type of therapy. After 6 weeks, the researchers measured participants’ fear of flying. The inter-esting part to this study was that the researchers again assessed participants’ fear of flying one year after the conclusion of the study. Results showed the levels of fear had remained relatively stable since the end of the study. Therefore, the researchers concluded that the introduc-tion of the independent variable (types of therapy) had caused a measurable change in the dependent variable (fear of flying). We mention this study to bring up the idea that researchers need to remain vigilant as to factors that might threaten the internal validity of their study. For example, do you think the follow-up results (i.e., results after 12 months) would have been different had
the events on September 11, 2001, occurred during that time? The answer is probably yes. Had the results been different and the fear level was higher, the researchers would not have been able to conclude that the independent variable had a lasting influence. There are many threats to internal validity that a researcher must be aware of when conducting research. We will discuss many of these threats later in Chapter 9. The next type of validity we want to discuss is external validity. External validity, also known as ecological validity, is the extent to which the obtained results in a study can be generalized to other settings. When considering external validity, you examine if any changes in the dependent variable can be applied to similar events. Specifically, can the results you obtained in the laboratory occur in a real-world setting? Researchers are often confronted with problems due to external validity. This is because a large percent-age of research is conducted in a laboratory environment, where the researcher can isolate a single independent variable to determine its influence on a dependent variable. However, since research is done in such controlled environments, it is sometimes diffi-cult to know if the causal inference drawn in the laboratory will apply to the real world. One way to combat threats to external validity is to design experiments as close to the real world as possible. For example, researchers in the area of cognitive psychology have done much research using microworld simulations and virtual computer games. Researchers continue to think outside of the box and design experiments that are as close to the real world as possible. This allows researchers to extend results obtained in the lab to the real world. The final type of validity that we will discuss is construct validity which is per-haps the most difficult to understand. Construct validity refers to the likelihood that the device or scale used to measure a variable actually is related to the topic or theory of interest. In other words, does the way we measure a variable accurately cap-ture the theoretical construct behind that variable? If we are interested in measuring college student obesity, we could ask the following questions: Do you overeat? Not at all Construct Validity: The likelihood that the device or scale used to measure a variable actually is related to the topic or theory of interest. External Validity: The extent to which the obtained results in a study can be generalized to other settings. 1 1 2 2 3 How often do you eat fast food? Not at all 3 Do you eat vegetables? Not at all 1 2 3 4 4 4 5 5 5 6 6 6 7 7 7 All the time All the time All the time On the surface, these questions might appear as though they will measure the variable of obesity. (This is known as face validity, where the device or scale has the superficial look to assess a variable’s theoretical construct.) However, what is the likelihood that participants in your study would be truthful in their answers? In addition, just because you eat a lot of food or fast food does not necessarily mean that you are obese. Con-versely, eating lots of vegetables does not mean you are skinny. If these were the ques- -tions you used to assess college student obesity, you might have low construct validity.
In order to have higher construct validity, you can assess two different components related to construct validity. The first component is convergent validity. The logic behind convergent validity is that your measure should converge or be similar to other measures of the same variable. Therefore, to have high construct validity your measure for a variable should show similar results to other valid measures of the same vari-able. Going back to the example on obesity, your measure should yield results simi-lar to that of a valid measure of obesity. The second component related to construct validity is divergent or discriminant validity. This is the opposite of convergent valid-ity. Whereas convergent validity argues your measure of a variable should be similar to other valid measures of the same variable, divergent validity argues your measure should be dissimilar to measures of different variables. Looking at the example of obe-sity again, your results from your measure should look similar to other obesity mea-sures, but should not look similar to say a measure on diabetes. By using these two related components when developing measures of variables, researchers are able to increase construct validity in their study. SECTION SUMMARY • Reliability is the consistency of your measure to produce similar results on dif-ferent occasions. There are several ways to assess reliability: • Test–retest reliability is a reliability assessment where your measure is tested on two different occasions for consistency. • Split-half reliability is a reliability assessment in which a measure is split in half and the two halves are compared. If the correlation is high, the measure is said to have high reliability. • Internal consistency is a reliability assessment similar to a split-half method. However, the splitting occurs more than once and an average of the correla-tions is taken. • The parallel-forms method is a reliability assessment in which a measure is divided in half and given to two groups of people. The reliability is high if each measure given is highly correlated. • Interrater reliability is used when a research design calls for observations of an event. Two or more observers compare results from their observations. The higher the observer consensus, the higher the reliability. • Validity is the accuracy of a measure to evaluate what it is supposed to measure. The three most common types of validity are internal, external, and construct validity. • Internal validity provides confidence in saying that the observed change in the dependent variable is due to the independent variable and not due to any outside influences. • External validity is the extent to which the obtained results in a study can be generalized to other settings.
• Construct validity refers to the likelihood that the device or scale used to mea-sure a variable actually is related to the topic or theory of interest. Construct validity is composed of two components, convergent validity and divergent validity. • Convergent validity states your measure should converge (or be similar) to other measures of the same variable. • Divergent validity argues your measure should be dissimilar to measures of different variables.
ARE YOU EQUIPPED NOW? Let’s revisit the topic of texting from the begin-ning of the chapter. This time we will pose the question, do you ever text and drive? Again, we have probably all done this before. However, is this safe to do? Research suggests it is not. For the purposes of this exercise, we are going to give you a modified version of an experiment conducted by Owens, McLaughlin, and Sudweeks (2011). Owens et al. wanted to manipulate texting conditions while driving (i.e., driving with no texting, driving while texting on a personal phone, and driving while texting using an in-vehicle texting system) to see if this influenced visual and steering behaviors of drivers. Participants completed these conditions by texting the researcher on a closed course. Results indicated that driving with no texting produced the best results, followed by the in-vehicle system and lastly the per-sonal phone. Thus, texting does reduce performance and is a mental distraction for drivers. For this experiment, answer the following questions: 1. What is the independent variable? 2. What is the dependent variable? 3. How did the researchers try to increase the ecological validity of the experiment?