bcs.worthpublishers.combcs.worthpublishers.com/.../irm/word/passer_ir_ch08.docxweb...

25
Chapter 8: Single-Factor Experimental Designs A. LEARNING OUTCOMES. After studying this chapter students should be able to: Describe the basic components of experimental control. Discuss how experimental control helps researchers satisfy three key criteria for inferring cause and effect. Describe ways to manipulate an independent variable. Discuss factors that affect the number of conditions incorporated into an experiment. Explain the concepts of experimental and control groups. Explain the advantages and disadvantages of between-subjects designs. Describe several types of between-subjects designs. Discuss the differences between random assignment and random sampling. Describe some advantages and disadvantages of within-subjects designs. Explain the goals of counterbalancing. Describe several types of within-subjects designs. Describe how descriptive statistics and inferential statistics help researchers examine the results of their studies. Discuss a typical approach used in single-factor designs to examine whether the findings are statistically significant. B.KEYWORDS All-possible-orders design Matched-groups design Between-subjects design Matching variable Block randomization Natural-groups design Block-randomization design Order effects Carryover effects Progressive effects Confounding variable Random assignment Control condition/control group Random-selected-orders design 85

Upload: trinhnhi

Post on 09-Apr-2018

223 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

Chapter 8: Single-Factor Experimental Designs

A. LEARNING OUTCOMES. After studying this chapter students should be able to: Describe the basic components of experimental control. Discuss how experimental control helps researchers satisfy three key criteria for inferring cause

and effect. Describe ways to manipulate an independent variable. Discuss factors that affect the number of conditions incorporated into an experiment. Explain the concepts of experimental and control groups. Explain the advantages and disadvantages of between-subjects designs. Describe several types of between-subjects designs. Discuss the differences between random assignment and random sampling. Describe some advantages and disadvantages of within-subjects designs. Explain the goals of counterbalancing. Describe several types of within-subjects designs. Describe how descriptive statistics and inferential statistics help researchers examine the results

of their studies. Discuss a typical approach used in single-factor designs to examine whether the findings are

statistically significant.

B. KEYWORDSAll-possible-orders design Matched-groups designBetween-subjects design Matching variableBlock randomization Natural-groups designBlock-randomization design Order effectsCarryover effects Progressive effectsConfounding variable Random assignmentControl condition/control group Random-selected-orders designCounterbalancing Reverse-counterbalancing designExperimental control Single-factor designExperimental condition/experimental group Subject variableIndependent-groups design Within-subjects designLatin Square

85

Page 2: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

86 CHAPTER 8: Single-Factor Experimental Designs

C. BRIEF CHAPTER OUTLINE

I. The Logic of ExperimentationA. Exercising Control over VariablesB. Causal Inference and Experimental Control

II. Manipulating Independent VariablesA. Varying the Amount or Type of a FactorB. Determining the Number of ConditionsC. Examining Nonlinear EffectsD. Experimental and Control Conditions

III. Between-Subjects DesignsA. Advantages of Between-Subjects DesignsB. Disadvantages of Between-Subjects DesignsC. Types of Between-Subjects DesignsD. Random Assignment Versus Random Sampling

IV. Within-Subjects DesignsA. Advantages of Within-Subjects DesignsB. Disadvantages of Within-Subjects DesignsC. The Need for CounterbalancingD. Types of Within-Subjects Designs

V. Examining the Results: General Concepts

D. EXTENDED CHAPTER OUTLINE*Much of this summary is taken verbatim from the text.

IntroductionThe chapter describes experiments in which a single variable is manipulated to create at least two

treatment conditions. Between-groups and within-groups single-factor designs are described as well as the advantages and disadvantages of each.

Part I: The Logic of ExperimentationIn an experiment, an independent variable, or factor of interest, is manipulated to determine its

effect on a dependent variable, or behavior. Because the experimenter keeps all other variables that might affect behavior constant in an experiment, a clear causal relationship between the independent variable and dependent variable can be established.

Page 3: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 87

A. Exercising control over variables. Experimental control enables researchers to conclude that the independent variable is the cause of any observed change in behavior. Experimental control is the ability to:a. manipulate one or more independent variables,b. choose the types of dependent variables that will be measured, and how and when they will be

measured, so that the effects of the independent variables can be assessed, andc. regulate other aspects of the research environment, including the manner in which participants

are exposed to the various conditions in the experiment. B. Causal inference and experimental control. In order to conclude that variable X has a causal effect

on variable Y, three criteria must be established. The methods used in experiments to establish these criteria, and how they meet the criteria are shown in the table below.

Criteria required for causal inference

Methods of experimental control

How experimental control meets criteria for causal inference

Covariation Systematically manipulating X A change in Y will only occur when X is manipulated.

Temporal order Exposing participants to X prior to any changes in Y

Y will only change when X is presented first.

Absence of alternative explanations

Holding constant other variables that could potentially affect Y

Eliminates those factors as potential explanations for Y varies.

A confounding variable is a factor that is not held constant in an experiment and that covaries with the independent variable in such a way that we can no longer determine which one has caused the change in the dependent variable. To reduce confounding factors, experimenters ( 1) keep extraneous variables as constant as possible across all treatment conditions, or (2) balance extraneous factors that , in principal, cannot be held constant. Potential confounding variables include environmental factors, such as a room’s temperature, humidity, and noise level, and the time of day when one participates as well as the experimenter himself or herself. Participant characteristics may also be confounding variables. Experimenters control for participant characteristics in between-subjects designs by randomly assigning participants to the various treatment conditions. In within-subjects designs participant characteristics are less of an issue because each participant engages in all experimental treatments. One issue with a within-subjects design, however, is the order in which participants engage in the levels of the independent variable. To reduce order effects, researchers use counterbalancing, a procedure in which the order of conditions is varied so that no condition has an overall advantage relative to the other conditions.

Part II: Manipulating Independent VariablesThis section describes the types of variables that are manipulated in order to determine their effect

on behavior, as well as what one should consider when deciding upon how many levels of the independent variable to create.

Page 4: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

88 CHAPTER 8: Single-Factor Experimental Designs

A. Varying the amount or type of a factor. In an experiment, treatment conditions are created by manipulating an independent variable so that there are at least two levels of it. In quantitative manipulations the independent variable is manipulated in amount (0 mg, 10 mg) whereas in qualitative manipulations the independent variable is manipulated based on type (e.g., red, blue, yellow). Examples of factors that are manipulated so as to create two or more treatment conditions include: the physical environment, social environment, interventions, tasks, strategies, and an organism’s characteristics.

B. A single-factor design is one in which a single independent variable is manipulated into two or more levels. When determining the number of conditions, researchers must consider the research question asked, their personal preferences, and available resources.

C. Examining nonlinear effects. In a linear relationship there is a constant change in Y for every change in X. Not all relationships between variables produce a perfect straight line, however. For example, both low levels of arousal and high levels of arousal are associated with impaired decision making, but moderate levels of arousal tend to produce accurate decisions. If a researcher manipulated arousal so there were only two levels, low and high, the effect of arousal on decision making would not be observed. By creating three or more levels of an independent variable the researcher is able to detect nonlinear relationships as well as provide more thorough data about linear relations.

D. In many experiments, comparisons are made between experimental and control conditions. An experimental condition involves exposing participants to a treatment, or an “active” level of the independent variable. In a control condition, participants do not receive the treatment of interest or are exposed to a baseline level of an independent variable. Sometimes an independent variable is manipulated to create the presence (experimental group) and absence (control group) of a factor; in other experiments the concept of control groups does not apply.

Part III: Between-Subjects DesignsBetween-subjects designs are those in which each participant engages in only one treatment

condition.

A. Some advantages of between-subjects designs are that effects produced by one condition do not carry over to other conditions, and that participants will not be “tipped off” about a treatment based on their participation in a previous experiment. Some experiments can only be conducted using a between-subjects design. Researchers examining first impressions is an example of such a study. After a person meets another person there is no possible way for a second first impression to occur.

B. Disadvantages of between-subjects designs are that the participants in each treatment condition are less likely to be equal and that the designs require a great number of participants.

C. Types of between-subjects designsa. In an independent-groups design, participants are randomly assigned to a treatment condition.

Random assignment does not ensure that treatment groups will be equal, but it does help in distributing differences across the treatment conditions in an unbiased way. Random assignment can be done through simple random sampling but that is not always ideal. In block randomization, participants are run in blocks of treatment conditions and blocks are repeated until the researcher obtains the number of participants needed.

Page 5: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 89

b. In matched-group designs participants are not assigned to treatment conditions on a purely random basis. Instead, participants are paired according to some characteristic or matching variable. Each member of every pair is then randomly assigned to one of the treatment conditions, thus creating, ideally, equivalent treatment conditions.

c. Natural-groups designs are those in which treatment conditions are based upon some naturally occurring characteristic. Such characteristics are called subject variables and include things such as gender, age, and extraversion. Subject variables are not true independent variables because they cannot be directly manipulated by the researcher. That said, the question becomes: is a natural-groups design an experiment? Most researchers refer to subject variables as independent variables, although some are more technical, and refer to them as “manipulated independent variables” or “quasi-independent variables.” Regardless of what one calls a subject variable, a research study that creates treatment conditions based on subject variables is not an experiment, but at best a correlation.

D. Random assignment and random sampling both rely on the laws of probability and are good scientific practice, but they are two different things. In random sampling every member of the population has an equal and independent chance of being selected into a sample. The goal of random sampling is to select a group of individuals whose characteristics are representative of the population. In contrast, random assignment is used to determine which treatment condition each member of the sample will engage in. In random assignment each member of the sample has an equal and independent chance of participating in any of the experiment’s treatment conditions. The goal of random assignment is to create treatment conditions that are unbiased so that when the study starts, there is no systematic difference between the treatment conditions that could contribute to whether there was a change in behavior.

Part IV: Within-Subjects DesignsA. In a within-subjects design each participant engages in all levels of the independent variable. The

key advantage of this design is that it requires significantly fewer participants as compared to the between-subjects design. Similarly, if one recruits the same number of participants in within-subjects design as in a between-subjects design the amount of data collected in the within-subjects design will be substantially greater than that from the between-subjects design. Finally, like the between-subjects design, there are some research questions that can only be answered using a within-subjects design. An example of one such study is one that examines individual perceptual thresholds. Disadvantages of within-subjects design are that participating in multiple levels of the independent variable will make people aware of the experiment’s purpose or hypothesis. Another limitation of the within-subjects design is the potential for order effects, which occur when participants’ responses are affected by the order of conditions to which they are exposed. Progressive effects reflect changes in participants’ responses due to their cumulative participation in the experiment. In other words, their participation in treatment A and treatment B have additive effects on how they will behave in treatment C. In contrast, carryover effects are when a person’s responses in a treatment are affected only by the last treatment in which they engaged.

Page 6: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

90 CHAPTER 8: Single-Factor Experimental Designs

B. Counterbalancing is necessary to reduce the potential for order effects. Studies that use counterbalancing still require their participants to engage in all levels of the independent variable, but the order in which they engage in each level differs from participant to participant. In this way no one has an advantage, or disadvantage, based on his or her position in the sequence.

C. There are several types of within-subjects designs.a. Those that expose participants to each condition once include:

i. In the all-possible-orders design there are n! (n factorial) unique orders that can be arranged, with n = the number of treatment conditions. For example, if there are four levels of an independent variable, then 4 x 3 x 2 x 1 = 24 possible orders. This complete counterbalancing technique achieves the three goals of counterbalancing, which are that (1) every condition of the independent variable appears equally often in each position, (2) every condition appears equally often before and after every other condition, and (3) every condition appears equally before and after every other condition within each pair of positions in the overall sequence. The key disadvantage of the all-possible-orders design is that the number of participants required grows exponentially larger as the number of levels of the independent variable increase.

ii. A Latin square design has an n (number positions in a series) X n (number of orders) matrix in which each condition will appear only once in each column and each row. For example, in an experiment where there are four levels of an independent variable, the number of treatment orders would be 4 x 4 = 16. The Latin square design achieves the first two, and most important, goals of counterbalancing that the all possible orders counterbalancing achieves (see above). The chief limitation of the Latin square design is that when there are an odd number of levels of the independent variable the second goal of counterbalancing is not achieved.

iii. In a random-selected-orders design a subset of orders are randomly selected from the entire set of possible orders. Each randomly selected order is then administered to one participant. This method of counterbalancing is likely to be used when the number of treatment conditions is particularly large or if the number of orders greatly exceeds the number of participants available.

b. Within-subjects designs can also expose participants to each condition more than once. This may be due to practical reasons, to examine the reliability of participants’ responses or to extend the generalizability of the results.i. In a block-randomization design every participant is exposed to multiple blocks of trials, with

each block for each participant containing a newly randomized order of all the conditions. For example, participant 1 will receive several blocks, each of which has a unique ordering of the levels of the independent variable. The major disadvantage of this design is that a single participant may engage in a great number of trials.

ii. In a reverse-counterbalancing design each participant receives a random order of all the conditions, and then receives them again in the reverse order. The logic of this design is that every level of the independent variable will have the same average order. A potential problem, however, is that there are nonlinear order effects. This occurs when the effect of receiving a level of the independent variable earlier (or later) does not balance out the effect of it being presented later (or earlier).

Page 7: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 91

Part V: Examining the Results: General Concepts The results of one-factor experimental designs are examined with descriptive statistics as well as

with inferential statistics. In descriptive statistics, researchers present a measure of central tendency and variability (usually the mean and standard deviation) for each treatment condition. Inferential statistics are used to determine whether an experiment’s results are statistically significant. Most experiments are designed to collect data at the interval and ratio level. When an experiment has only two levels of an independent variable, a t-test is used to determine whether the mean score of each treatment condition is statistically different from the other. When there are more than two levels of an independent variable, an analysis of variance (ANOVA) is performed to detect differences between group means. ANOVA only tells the researcher whether a difference exists, not where the difference lies. As such, if an ANOVA reveals a statistically significant effect, another analysis, namely a post-hoc test must be performed to determine which groups differ from one another.

C. LECTURE AND CLASSROOM ENHANCEMENTS

PART I: The Logic of Experimentation

A. Lecture/Discussion Topics The probability of causal relations among variables. The research methods described thus far have

all been nonexperimental and because they are nonexperimental, they can never make causal statements about relations between variables. The experiment is the ONLY research method capable of making causal statements about independent variables and dependent variables due to the high level of control exerted in their design. However, just because a research study claims to have performed an experiment does not necessarily mean that the variables examined are causally related. While the experiment is often described as having the ability to establish causal relationships, it is critical for consumers of research to understand that the causality between two variables is still based on probability. Experimental data are subjected to parametric statistical tests to determine how probable it is that the independent variable caused the change in the dependent variable. When an experiment’s results suggest a statistically significant causal relationship between X and Y it means that it is statistically more probable that not that the variables are causally related. However, there is always the possibility that X and Y are not related whatsoever and that the observation of their being related occurred out of pure chance. Statistical analyses enable the researcher to set parameters (alpha level) to reduce the error of making false statements about research results (type I and type II error).

Page 8: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

92 CHAPTER 8: Single-Factor Experimental Designs

Independent variables, subject variables, experiments, and quasi-experiments. Subject variables are introduced later in the chapter but it’s not a bad idea to introduce the term early on when talking about experiments. Many research studies are published as experiments, and even use the word experiment in their title when, in fact, they are not true experiments. It is important for students to understand that an independent variable is something that is manipulated directly by the researcher. In other words, if the variable is something that does vary, but through a way other than the researcher’s direct involvement, it is not a true independent variable. The classic example of a variable often referred to as an independent variable is sex. A researcher can examine the differences between men and women, but the levels of sex were in no possible way manipulated by the researcher; maleness and femaleness is something that varies naturally. Naturally occurring independent variables are technically referred to as subject variables but in many research studies they are still identified as the independent variable. The use of subject variables can never lead to statements of causality because, by nature, they lack experimental control.

B. Classroom Exercise Identifying alternative explanations and Exerting Experimental Control. The table below lists

independent variables and dependent variables that have been linked causally through experimental research. To give students practice in critical thinking, have them identify other factors that could have potentially led to the observed change in each dependent variable. Next, ask them to think of ways one could control the potential confounds they identified so as to ensure that the independent variable is the only likely reason for the change in the dependent variable.

Independent Variable Dependent Variable Potential Confounding Variables

Exercise Memory Diet, age, education level

Dance Therapy Body image The act of dancing, going to therapy

Smoking Weight gain Overeating, physical activity

Social support Job satisfaction Type of industry one is in, co-workers, salary

Personal space Anxiety Drug use, anxiety disorder diagnosis, culture

Page 9: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 93

Experiment or quasi-experiment? One way a researcher exerts experimental control in an experiment is by directly manipulating a variable to determine whether it causes a change in another variable. Although the basic design of quasi-experiments is almost identical to true experiments, because variable X is not directly manipulated, but is instead measured as it naturally exists, they cannot make casual statements about the relation between X and Y. To help students distinguish between experiments and quasi-experiments provide them with the following list of references. Ask them to determine, based on the study’s title, whether it is more likely a true experiment or a quasi-experiment, and to provide a rationale for their answer. Next, have them think of ways to turn the quasi-experimental studies into true experiments.

Lin, C., & Huang, M. (2011). Effects of trait oping styles and emotional display rules on emotional labor strategies. Acta Psychologica Sinica, 43(1), 65–73. (Quasi-experiment; trait coping style is a subject variable).

Ramchandani, P. G., O'Connor, T. G., Evans, J., Heron, J., Murray, L., & Stein, A. (2008). The effects of pre- and postnatal depression in fathers: A natural experiment comparing the effects of exposure to depression on offspring. Journal of Child Psychology and Psychiatry, 49(10), 1069-1078. doi:10.1111/j.1469-7610.2008.02000.x (quasi-experiment; fathers’ depression is a subject variable)

Snodin, N. S. (2013). The effects of blended learning with a CMS on the development of autonomous learning: A case study of different degrees of autonomy achieved by individual learners. Computers & Education, 61,209–216. doi:10.1016/j.compedu.2012.10.004. (experiment; learning style can be directly manipulated)

Yang, Q., Wu, X., Zhou, X., Mead, N. L., Vohs, K. D., & Baumeister, R. F. (2013). Diverging effects of clean versus dirty money on attitudes, values, and interpersonal behavior. Journal of Personality and Social Psychology, 104(3), 473–489. doi:10.1037/a0030596 (experiment; cleanliness of money can be directly manipulated)

C. Web Resources On causal relationships. http://www.utexas.edu/research/pair/causal.htm#1 A statistician’s perspective on causality.

http://www.maths.bris.ac.uk/~maxvd/Consilience_Did.pdf Research methods by dummies (experiments and variables).

http://psych.csufresno.edu/psy144/Content/Design/Types/experimental.html

D. Film Suggestion Experimental designs are discussed in the Against All Odds: Inside Statistics film series. Program

12 in the series provides examples of experimental research and describes the importance of random selection and assignment in creating treatment conditions. The video can be accessed at http://www.learner.org/vod/vod_window.html?pid=150

Consort for Mathematics and Its Applications, I., Amabile, T. M., Singer, J., & Intellimation, I. (1989). Against All Odds: Inside Statistics. Santa Barbara, CA: Intellimation.

Page 10: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

94 CHAPTER 8: Single-Factor Experimental Designs

E. Additional References. On causality in research.

Frosch, C. A., & Johnson-Laird, P. N. (2011). Is everyday causation deterministic or probabilistic? Acta Psychologica, 137(3), 280-291. doi:10.1016/j.actpsy.2011.01.015.

Shugan, S. M. (2007). Causality, unintended consequences and deducing shared causes. Marketing Science, 26(6), 731–741. doi:10.1287/mksc.1070.0338.

PART II: Manipulating Independent Variables

A. Lecture/Discussion Topics How different should the levels of an independent variable be? A simple experiment is one that

manipulates a single variable to determine whether its presence (experimental group) or absence (control group) affects behavior. But just how different should the experimental and control levels be to detect its effect on behavior? If we take exercise, for example, one can argue that walking from one end of the classroom to another is, technically speaking, exercise, but is it sufficient to produce a change in memory compared to a sedentary control group? Probably not. When creating treatment conditions the researcher must be cognizant that the strength of the manipulation is a critical factor that can affect whether she observes a change in behavior due to the independent variable. When the levels of the independent variable are too similar to one another, a difference in behavior across the treatment conditions is unlikely to be observed. On the other hand, if the manipulation is too strong, the study may also be negatively affected. For example, if one is studying the effect of a drug on behavior, giving too strong a dose could physically harm the participant. Some experiments, particularly those that involve drugs, are designed to examine dose-response curves. In this design there are many levels of the independent variable, each of which is differs in magnitude. A dose-response study provides information about the minimum dose necessary to produce a change in behavior as well as the dose at which toxicity is produced. Manipulation checks. Manipulation checks are discussed in Chapter 10 but they can also be

introduced while talking about the basic experimental design. Researchers would be wise to perform a manipulation check on their independent variable. If the independent variable is exercise, for example, the presence and absence of exercise in participants may be checked by measuring their heart rate before the study and during participation (while they are receiving the control or experimental treatment). This manipulation check provides assurance to the researcher that the treatment is creating the intended situation or environment. Sometimes a manipulation check is conducted after a person’s participation is over. For example, if participants were supposed to experience stress (or no stress), the researcher may ask participants to indicate the degree to which they experienced stress. This check should reveal that participants in the stress condition report greater levels of stress than those in the no-stress control group, which lends credibility, or validity, to the researcher’s operationism of stress.

Page 11: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 95

B. Classroom Exercise Qualitative and quantitative independent variables. To get students to understand how the levels

of independent variables can be qualitative or quantitative in nature, provide them with a general description of the purpose of several studies as well as the independent variable manipulated in each. Have them work on manipulating each independent variable so that the levels reflect (1) qualitative differences and (2) quantitative differences.

The purpose of the study is to

examine:

Independent Variable Possible Qualitative Levels

Possible Quantitative Levels

feedback on job performance

feedback positive, negative or neutral

none, a little, a lot

whether color affects mood

color red, blue, green or yellow

red in various saturation levels

whether diet alters the immune response

diet raw, Vegan, Carnivorous

protein from non-meat sources or protein predominantly from meat.

C. Web Resource Web link to a quiz on manipulation checks. After a person completes the five-item quiz he or she

has the option to view the correct answers. http://webquiz.ilrn.com/ilrn/quiz-public;jsessionid=8ED2A6A639F74F5062EDEB97430BD0C9?name=stmr01q%2Fstmr01q_WS_chp14&cookieTest=1

D. Additional References Experiments revealing nonlinear relationships.

Baldi, E., & Bucherelli, C. (2005). The inverted "u-shaped" dose-effect relationships in learning and memory: Modulation of arousal and consolidation. Nonlinearity in Biology, Toxicology, and Medicine, 3(1), 9–21

Viinikainen, M., Jääskeläinen, I. P., Alexandrov, Y., Balk, M. H., Autti, T., & Sams, M. (2010). Nonlinear relationship between emotional valence and brain activity: Evidence of separate negative and positive valence dimensions. Human Brain Mapping, 31(7), 1030–1040.

Page 12: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

96 CHAPTER 8: Single-Factor Experimental Designs

PART III: Between-Subjects Designs

A. Lecture/Discussion Topics Analyzing single-factor between-subjects designs. In a between-subjects design an independent

variable is manipulated so that there are two or more levels of it. Participants are then randomly assigned to engage in only one level of the independent variable. The type of data analysis performed on single-factor between-subjects designs depends on two things: (1) what level the dependent variable is measured at (nominal, ordinal, interval, or ratio), and (2) how many levels of the independent variable there are. Even though this may seem straight forward, it can be very overwhelming to undergraduates. Decision trees are a great tool to aid one in selecting the correct students in making decisions about what statistic to use. The following is a decision tree you can give to students to use when deciding which statistical test they should use. The document answers which test is appropriate for all research methods covered in this text. It can be downloaded at: http://www.muhlenberg.edu/pdf/main/academics/psychology/stats_decision.pdf

How random assignment can overcome (to a degree) nonrandom selection. Most research in psychology relies upon the use of a convenience sample. As discussed in Chapter 7, convenience samples are nonrandom samples that can introduce bias.

B. Classroom Exercises Deciding whether to use a between-groups or within-groups design. Since most experiments can

be approached using a between-groups design or a within-groups design, experimenters must decide which approach is best for a specific research question. Present students with the following research questions and have them produce two experimental designs for each:a between-groups a within-groups one. For each research question ask the students to identify at least two advantages of using a within-subjects design and two disadvantages of using a between-subjects design. This exercise can be used to help demonstrate how practice effects are an important consideration in research and also how within-subjects designs enable the researcher to use fewer participants.

o Do diets rich in antioxidants decrease dementia-like symptoms in aged mice? o Does exposure to an enriched environment affect long-term potentiation (LTP is believed to

be the cellular correlate of learning and memory)?o Does homework improve performance in statistics?

C. Web Resources Random selection and random assignment.

http://www.socialresearchmethods.net/kb/random.htm

D. Additional References Tips on generating hypotheses in experiments.

McGuire, W. J. (1997). Creative hypothesis generating in psychology: Some useful heuristics. Annual Review of Psychology, 48(1), 1–-30.

Page 13: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 97

PART IV: Within-Subjects Designs

A. Lecture/Discussion Topics Choosing the Best Research Method. How one conducts research is dependent upon the research

question he/she has asked. Present the class with the following research questions and have them identify which research method/s might best answer each. Be sure to have students provide rationale for their answers.

1. How does fetal alcohol exposure affect the cognitive development of children? 2. Does aerobic activity improve memory retention in adults?3. How to psychology majors feel about the Graduate Record Exam?4. What are the long-term effects of smoking cessation programs?

Research Methods Are Separate But Equal. Psychology is a very diverse discipline and the various subsets of it use different research methods. For example, whereas biological psychologists tend to conduct experimental laboratory research, social psychologists typically perform descriptive field studies. Although the different disciplines utilize different research methods, emphasize to the students that the overall goal of psychology is to understand human behavior and that different disciplines simply approach this goal from many different directions.

B. Classroom Exercise Generating Experimental Hypotheses from Non-Experiments. Many research methods lead to

future experiments. Provide students with the following brief research articles (or supply your own) and ask them to create a hypothesis based on the work that could be tested experimentally. This exercise will have students learn how the same research topic can be approached from multiple directions.

Harden, S. L., Clark, R. A., Johnson, W. B., & Larson, J. (2009). Cross‐gender mentorship in clinical psychology doctoral programs: an exploratory survey study. Mentoring & Tutoring: Partnership in Learning, 17(3), 277-290.

Ng, E. S., Schweitzer, L., & Lyons, S. T. (2010). New generation, great expectations: A field study of the millennial generation. Journal of Business and Psychology, 25(2), 281-292.

Tindell, D. R., & Bohlander, R. W. (2012). The use and abuse of cell phones and text messaging in the classroom: A survey of college students. College Teaching, 60(1), 1-9.

C. Web Resource Qualitative and Quantitative Research. This is a brief and entertaining clip that describes the

similarities and differences between quantitative and qualitative research http://www.xtranormal.com/watch/12153618/qualitative-vs-quantitatinve research

A Truly “Long”-itudinal Design. The Wisconsin Longitudinal Study (WLS) began in 1957 with the intent of understanding how experiences early in life may impact one’s quality of life later in life. This link describes the study and provides a timeline to put the work into perspective. http://www.news.wisc.edu/13941

Page 14: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

98 CHAPTER 8: Single-Factor Experimental Designs

D. Film Suggestion Understanding Research. The Annenberg Foundation provides free access to the film series,

Discovering Psychology. All films can be accessed online and played on demand. The second film in this series is “Understanding Research” and it describes the various ways in which psychological research is conducted both in the field, as well as in the lab. http://www.learner.org/series/discoveringpsychology/02/e02expand.html

Zimbardo, P. G. (1989). Discovering psychology. South Burlington, VT: Annenberg/CPB Collection.

E. Additional references On internal and external validity:

Anderson, C. A., Lindsay, J. J., & Bushman, B. J. (1999). Research in the Psychological Laboratory Truth or Trivia? Current directions in psychological science, 8(1), 3-9.

Onwuegbuzie, A. J., & McLean, J. E. (2003). Expanding the Framework of Internal and External Validity in Quantitative Research. Research in the Schools.

PART V: Examining the Results: General Concepts

A. Lecture/Discussion Topics Describing sample data. All research methods enable one to describe the data collected. The ways

in which data are described depend on the type of research method used as well as how the dependent variable was measured. In this particular chapter the experiment is discussed and, typically, but not always, data is collected at the interval or ratio level. Both levels of measurement allow the researcher to describe the samples based on a mean score. Other levels of measurement also provide the calculation of the “typical” score, but not as the arithmetic mean. Instead, data measured nominally are described by the mode, or most frequently occurring score, and data at the interval level are described at either the mode or the median, the middlemost value.

It’s statistics, not sadistics. If there is one thing my experimental psychology students have in common it’s their overwhelming fear of statistics. Even after completing a statistics prerequisite students still cringe at the word statistics. Rather than ignoring students’ anxiety around statistics, take the opportunity to help them understand that statistics really aren’t as bad as they perceive them to be. Psychologists are in the business of understanding human behavior. To that end, we examine many people across many different situations. In quantitative research we can take this behavior and describe it with a number e.g., mean). Using the mean score of a sample is nothing more than using a number, instead of a word, to describe how all of the people behaved, on average. It’s a way to describe, in one number, what is typical of many (sometimes thousands!) individuals. I find that discussing statistics in a natural way during conversations regarding research methods eases students’ anxiety as they realize that statistics are just another thing in the world of psychology.

Page 15: bcs.worthpublishers.combcs.worthpublishers.com/.../IRM/Word/Passer_ir_ch08.docxWeb viewbcs.worthpublishers.com

CHAPTER 8: Single-Factor Experimental Designs 99

B. Classroom Exercise Thinking statistically. If students will be conducting research projects, allow them time to think

about how they will analyze their data and what types of conclusions they will be able to draw from them. You may want to have them start by identifying what they will measure, how they will measure it, and what measure of central tendency the measure will provide.

C. Web Resource Statistics cheat sheet. This link is to a “cheat sheet” for basic descriptive and inferential statistics

that serves as a nice, brief review of the basics. http://www.dummies.com/how-to/content/statistics-for-dummies-cheat-sheet.html

D. Film Suggestion Introductory statistics. The Annenberg Foundation has a variety of films freely available to use in

the classroom. This particular series, Against All Odds: Inside Statistics, includes 26 separate videos on basic statistical methods. http://www.learner.org/resources/series65.html

E. Additional References Creating clear and accurate research reports.

Bakker and Wicherts (2011) examined the accuracy of statistics in psychology journal articles, finding numerous errors. This paper can be used to facilitate discussion in class about the use of statistics in psychology research and what impact misreporting of results can have on the field. Bakker, M. & Wicherts, J.M. (2011). The (mis)reporting of statistical results in psychology journals. Behavioral Research Methods, 43, 666–678.

Wilkenson, L. (1999). Statistical methods in psychology journals: Guidelines and explanations. American Psychologist, 54, 5–-604.