1. Small N designs study a small number of subjects, often using ABA reversal designs or multiple baseline designs. They are useful when few subjects are available or when studying clinical populations.
2. Large N designs compare groups of subjects and allow for inferential statistical analysis. However, they require a large number of subjects and do not consider individual responses.
3. Both small and large N designs use baseline conditions to measure behavior without experimental manipulation. Small N designs are often preferred in early operant conditioning research due to their ability to study individual behavior changes closely.
1. Small N designs study a small number of subjects, often using ABA reversal designs or multiple baseline designs. They are useful when few subjects are available or when studying clinical populations.
2. Large N designs compare groups of subjects and allow for inferential statistical analysis. However, they require a large number of subjects and do not consider individual responses.
3. Both small and large N designs use baseline conditions to measure behavior without experimental manipulation. Small N designs are often preferred in early operant conditioning research due to their ability to study individual behavior changes closely.
1. Small N designs study a small number of subjects, often using ABA reversal designs or multiple baseline designs. They are useful when few subjects are available or when studying clinical populations.
2. Large N designs compare groups of subjects and allow for inferential statistical analysis. However, they require a large number of subjects and do not consider individual responses.
3. Both small and large N designs use baseline conditions to measure behavior without experimental manipulation. Small N designs are often preferred in early operant conditioning research due to their ability to study individual behavior changes closely.
1. Small N designs study a small number of subjects, often using ABA reversal designs or multiple baseline designs. They are useful when few subjects are available or when studying clinical populations.
2. Large N designs compare groups of subjects and allow for inferential statistical analysis. However, they require a large number of subjects and do not consider individual responses.
3. Both small and large N designs use baseline conditions to measure behavior without experimental manipulation. Small N designs are often preferred in early operant conditioning research due to their ability to study individual behavior changes closely.
EXP FINALS REVIEWER test a treatment when there How did Kazdin explain the
are insufficient subjects to decision of many clinical
Within subject designs small conduct a large N study and researchers to end without a N when she wants to avoid the return to baseline? What is a large n design? ethical problem of an It would be ethically untreated control group. A large N design compares indefensible to cause a the performance of groups of how about animal patient to relapse by subjects. researcher? returning to baseline after treatment appeared to How does a large N design Animal researchers prefer improve behavior. differ from a small N design? small N designs to minimize the acquisition and When is this most A small N design studies one maintenance cost, training important? or two subjects, often using time, and possible sacrifice of variations of the ABA reversal When relapse threatens the their animal subjects. design (A = baseline health or safety of the condition; B = experimental Which historical patient or others, as in in condition; A = verify the development caused the self-injurious, and suicidal or change in behavior). shift to large N design? homicidal behavior
Aggregate effects Sir Ronald Fisher’s (1935) What price do researchers
creation of the analysis of pay when they can’t return Aggregate effects are the variance allowed inferential to baseline? pooled findings from many testing of large N data. subjects They can’t rule out the Where have small N designs possibility that the patient’s Why do small N research been most extensively used? clinical improvement was challenge large N caused by an extraneous experiments? Small N designs have been variable. most extensively used in They argue that large N operant conditioning Multiple baseline design studies ignore individual research. subject responses to the IV In a multiple baseline design, and instead report aggregate B. F. Skinner examined the a series of baselines and results or trends. When continuous behavior of treatments are compared subjects vary greatly in their individual subjects in within the same subject, and response to the IV, this can preference to analyzing once treatments are create the appearance of no discrete measurements from administered, they are not difference between the separate groups of subjects. withdrawn. groups. Baseline This approach could also be Why might a clinical used to evaluate the effect of In both large and small N psychologist use small N a treatment administered to designs, baseline are control designs? different individuals after conditions that allow us to baselines of different A clinical psychologist could measure behavior without lengths. A researcher can use a small N design to the influence of the IV. evaluate the effects of a treatment on two or more participants succeed. For The similarity of human behaviors or on the same example, initially, a subject sensory systems allows behavior in different settings. might receive a reward for 30 researchers to generalize minutes of daily exercise, from a small number of How might this design later, for 45 minutes, and subjects. overcome the ethical finally, for 60 minutes. problem of withdrawing an When is a small N designs effective treatment? Reinforcement for successive appropriate? approximations of the target In a multiple baseline design, When studying a clinical behavior is central to athletic an experimenter never subject (a self-injurious child) training, behavior withdraws treatments after or when very few subjects modification, and administering them. are available. biofeedback and neuro How do researchers analyze feedback. When would we prefer a data from small N large N design? Discrete trials design experiments? A large N design would be A discrete trials design is a Researchers often visually desirable when we have small N design without inspect changes in the sufficient subjects and want baselines used in dependent variable across to increase generalizability. psychophysical research. treatment conditions. The Instead, the impact of The generalizability of a large independent variable’s effect different levels of the N study depends on how we is often apparent. independent variable is select our sample since a They may also use statistics averaged across 100s to seriously biased sample will to analyze small N data. 1000s of trials. not represent the population. Why is statistical analysis of How does a discrete trials small N data controversial? design differ from a typical Why doesn’t a large N study experiment? always have greater Critics are concerned about generalizability than a small generalizing from a single A discrete trials design has N study? subject to a population. no baselines and administers the levels of the independent If a large N study’s sample is Unless 50 measurements are variable 100s to 1000s of biased, we will be unable to taken during each baseline times to each subject. generalize its findings to a and treatment phase, larger population. Also, if it is important assumptions What are a discrete trails poorly controlled, there will underlying inferential tests designs benefits? be no valid findings to may be violated. The large number of data generalize. Changing criterion designs points produced by 100s to Why doesn’t a large N study 1000s of trials provides a In changing criterion designs, always have greater very reliable measurement of the criteria for reinforcement generality a small N study? the effect of the independent are incrementally increased variable. In contrast, a well-controlled as small N experiment using a single subject might share at least one What is statistical be successfully replicated characteristic in common significance? across sufficient subjects to (like passengers on a car). Results are statistically generalize its results to the A sample is a subset of the significant when the population from which they population that we use to difference between our were drawn. draw inferences about the treatment groups exceeds Why we need statistics? population. the normal variability of scores on the dependent The Argument for Using What does it mean when we variable. Statistics conclude that our scores probably came from same Statistical significance means Statistics are quantitative population? that there is a treatment measurements of samples. effect at an alpha level we Through statistics we can The differences in scores have preselected, like .01 or quantify the phenomena we obtained from separate .05. observe. It provides treatment groups are not researchers to with objective significantly greater than Alternative hypothesis and consensual techniques what we might expect The alternative hypothesis for describing their results. between any samples (H1) is the statement that randomly drawn from this What do statistics tell us? the scores came from population. When different populations the Statistical Inference is about researchers report this independent variable making a statement about outcome, it means that were significantly affected the the population and all its was no treatment effect. dependent variable. samples based on what we What is variability? see in the data that we have. When may we reject the null For a set of dependent hypothesis? Descriptive statistics variable measurements, describe sample central We may reject the null there is variability when the tendency and variability. hypothesis when the scores are different. differences between Inferential statistics allow us Variability “spreads out” a treatment groups exceed the to draw conclusions about a sample of scores drawn from normal variability in the parent population from a a population. dependent variable at our sample. In research, we do chosen level of significance. not outright conclude a What is null hypothesis? statement; but rather proves What does a frequency The null hypothesis (H0) is a certain probability. distribution of scores the statement that the reveal? Define sample and scores came from the same population population and the The frequency distribution independent variable did not displays the number of A population is a set of significantly affect the individuals contributing a people, animals, or objects dependent variable. specific value of the that dependent variable in a sample. What does a frequency A non-directional hypothesis How should we support null distribution of score reveal? predicts that the two groups hypothesis testing? will have different values on The values of the dependent An American Psychological the dependent variable: variable are indicated Association task force For example: The recommended that on the horizontal X-axis experimental group and researchers include (abscissa) and the control estimates of effect size and frequencies of these values confidence intervals, in are indicated on the vertical group will achieve different addition to p values. Y-axis (ordinate). You can systolic blood pressure calculate the total number of reductions. When you calculate a p value participants by adding the that is statistically significant, What is a significance level frequencies. this means that your results and how do we select one? are unlikely to be due to Why does rejecting the null The significance level (alpha) chance (are probably real). hypothesis depend on data is our criterion for deciding variability? Effect size whether to accept or reject The decision to accept or the null hypothesis. Effect size estimates the reject the null hypothesis strength of the association Psychologists do not use a depends on whether the between the independent significance level larger than differences we measure and dependent variable— .05. between treatment groups the percentage of the are significantly greater than What is significance level variability in the dependent the normal variability among and how do we select one? variable is due to the people in the population. independent variable. A significance level of .05 The greater the normal means that a pattern of How should we support null variability in the population, results is so unlikely that it hypothesis testing? the larger the difference could have occurred by A confidence interval is a between groups required to chance fewer than 5 times range of values above and reject the null hypothesis. out of 100. below a sample mean that is Contrast directional and What are type 1 and type 2 likely to contain the non-directional hypothesis errors? population mean (usually 95% or 99% of the time). A directional hypothesis A Type 1 error (a) is rejecting predicts the “direction” of the null hypothesis when it is What is critical region? the difference between two correct. The experimenter A critical region is a region of groups on the dependent determines the risk of a Type the distribution of a test variable. 1 error by selecting the alpha statistic sufficiently extreme level. For example: The to reject the null hypothesis. experimental group will A Type 2 error (b) is For example, if our criterion lower their systolic blood accepting the null hypothesis is the .05 level, the critical pressure more than the when it is false. region consists of the most control group. extreme 5% of the What is the level of an absolute zero. This scale distribution. measurement of the allows us to state that 2 dependent variable meters are twice as long as 1 To reject the null hypothesis, Are the subjects meter. the test statistic would have matched? to fall within the shaded Example: distance in meters critical region. Summarize the properties of or time in seconds. the four levels of What are the one tailed and How does the level of measurement two tailed test? measurement affect the test A nominal scale assigns tines we sue? A one-tailed test has a to two or more distinct critical region at one tail of Nonparametric tests use categories that can be the distribution. We use a nominal or ordinal data. named using a shared one-tailed test with a feature, but does not Parametric tests require directional hypothesis. quantify items. interval or ratio data. A two-tailed test has two Example: you can sort When should we use the chi critical regions, found at pictures into attractive and square test? opposite ends of the unattractive categories. distribution. We use a two- When the data are nominal tailed test with a non- An ordinal scale measures and the groups are directional hypothesis. the magnitude of the DV independent, which means using ranks. This scale allows the experimenter assigns What is the function of us to make statements about different subjects to them. inferential statistics? contestants’ relative speed. What does it reveal? Inferential statistics allow us Example: marathon to predict the behavior of a The chi square test contestants are assigned to population from a sample. determines whether the places from first place to last frequency of sample Examples of inferential place. responses represents the statistics are the t test and F An interval scale measures frequencies we would expect test. the magnitude of the DV in the population. Analyzing results using equal intervals Explain the meaning of the between values with no The parameters of data obtained (C2 obt) and critical absolute zero point. analysis values, and cramer’s Example: Fahrenheit or coefficient delta? How many Centigrade temperatures, independent The c2 obt is the actual and Sarnoff and Zimbardo’s variables are there? frequency of responses. 0-100 scale. How many treatment The critical value is the conditions are there? A ratio scale measures the minimum value required to Is the experiment magnitude of the DV using reject the null hypothesis. run between or equal intervals between within subjects? values and Cramer’s coefficient Φ is When may we reject the null When do we use an analysis analogous to r2 and indexes hypothesis? of variance? the degree of association We reject the null hypothesis We use an analysis of between priming and the when tobt > tcrit. For 9 df, if variance when data are number of incorrect tobt > 2.262, we would reject interval or ratio level and responses. the null hypothesis. there is at least one If our sample included every independent variable with Calculate an effect size for a member of the population, three or more levels. t test for independent we would have the maximum groups. What are within-groups possible degrees of freedom variability and between- and would know the exact First, we calculate the t groups variability? population values of the statistic (2.47) and then we mean and variance. enter it into the following Within-groups variability is formula: the degree to which the How do we decide whether scores of subjects in the to accept or reject the null same treatment group differ hypothesis? from each other. If x2 obt > x2 critical, reject Within-groups variability the null hypothesis. consists of error due to How does sample size affect individual differences and the t-test? extraneous variables.
The sample size determines Between-groups variability is
the degrees of freedom What is a t test for matched the degree to which the groups? scores of different treatment There is a different t groups differ from one distribution for each value of A t test for matched groups another or the grand mean. degrees of freedom. either assigns the same subjects to both conditions Between-groups variability How does t distribution or matches subjects and then consists of error due to change as sample size randomly assigns them to individual differences and increases? either condition. extraneous variables and The t distribution approaches treatment effects. A t test for matched groups a normal curve as sample may use fewer subjects and What does it mean when an size increases. achieve greater control over F ratio is statistically What does robustness individual differences than a significant? mean? t test for independent Across all group means, groups. The t test provides a valid there is a significant test of the hypothesis when This makes a t test for difference due to the assumptions like normal matched groups potentially independent variable. distribution of population more powerful. values are slightly to moderately violated. We compare the relative The maximum number of An internally valid sizes of the components of comparisons = p – 1, where p experiment is free of ANOVA by computing a ratio is the number of treatment confounding. between them groups. What is a manipulation What is the advantage of a check? priori test over a post hoc A manipulation check test? evaluates how well the A priori tests are more experimenter manipulated powerful than post hoc tests; the experimental situation. but you may perform fewer a A manipulation check priori tests. determines whether subjects Effect size followed directions and were When is post hoc test appropriately affected by our appropriate? Effect size measured by η2 is treatments. the proportion of the When an overall ANOVA is variability in the dependent What did Orne (1969) mean significant and you have variable that can be by a pact of ignorance? made no specific predictions, accounted for by the you may perform Subjects expect their data to independent variable. be discarded if they guess post hoc tests on all pairs of η2 indexes the strength of the experimental hypothesis, treatment groups. This is a the relationship between the and don’t volunteer this test done after the overall independent and dependent information to the analysis indicates a variables. experimenter. significant difference. Note: Debrief, to counter How many comparisons may this, subjects need to believe we perform? that you really need to You may perform all possible believe that you really want a pairwise comparisons truthful answer. without increasing the risk of Experimenters don’t want to Type 1 error. Drawing conclusions, the test additional subjects and When may we use a prior search for the elusive may take subject reports at test? bottom line face value instead of requesting additional We may use a priori tests to What is internal validity? information. test predictions of An experiment is internally differences between groups, Note: when you interpret, valid when the effects on the always be open to other such as between two groups dependent variable are due or between one group and potential explanation for to the independent variable. your findings. the others. Which mistakes could Generalizing across subjects We cannot be sure of the produce this threat to is critical to the external reliability or validity of our internal validity? validity and usefulness of procedures. experimental findings. Selecting the wrong What is research statistical test Using a t-test Which problems prevent us significance? to analyze ordinal data. from generalizing across A study achieves research subjects? Improperly using a statistical significance when its findings test Calculating multiple t- The samples used in clarify or extend knowledge tests. psychological research are gained from previous studies often biased and may not and raise implications for Drawing the wrong represent the larger broader theoretical issues. conclusions from the test population. Reporting p = .07 as a trend. When should we question The samples may not always novel findings? Statistical Conclusion represent even college Validity – the validity of We should question novel sophomores since we heavily drawing conclusions about a findings when they depend on volunteers. treatment effect from the contradict prior findings that statistical results that were Note: volunteers can be have been successfully obtained. quite different from non- replicated. volunteers. External Validity The burden of proof is on the Explain generalizing from experimenter who claims An experiment is externally procedures to concepts novel findings to explain this valid when its findings can be discrepancy. extended to other situations Experimental variables like and populations. anger may have multiple Why do we want to operational definitions. generalize beyond the What two requirements laboratory? must an externally valid When we generalize from study satisfy? our experimental results, we We want to generalize move from discussing our beyond the laboratory to 1. The experiment must be specific operational increase the external validity internally valid. definition of anger to of our findings. 2. The experimental findings discussing the concept of What is the problem in can be replicated. anger itself. extending laboratory What does it mean to Why is this a problem in findings? generalize across subject? research? Since extraneous variables why is this important? It is dangerous to generalize are uncontrolled in real The findings can be extended from a single world setting and operate in to a larger group than our complex combinations, they experiment’s operational sample. can modify the influence of definition of anger. our individual variables. What is the trade-off of data to increase external What is the advantage of between laboratory and validity. multiverse design? field experiments? Combining the results of Multivariate designs allow us The trade-off is between the experiments with different to study the effect of an laboratory’s more precise subjects and methodologies independent variable on control of extraneous increases the generality and combinations of dependent variables and the field external validity of our variables. experiment’s greater realism findings. These designs better and external validity. How is this logic shared by simulate the complexity of What did Hanson (1980) find meta-analysis? the real world than about the agreement univariate designs and Meta-analysis uses statistical between laboratory and provide more detailed analysis to combine and field studies? information. quantify data from many Hanson (1980) found that comparable experiments to How are multivariate more laboratory than field calculate an average effect experiments analyzed? studies reported a positive size. We analyze multivariate correlation between How does aggregation experiments with a reported attitudes and establish external validity? multivariate analysis of behavior. variance (MANOVA). Aggregation establishes When can we extend results external validity by How should researchers from an experiment to combining the results of handle a non-significant everyday life? experiments performed outcome? We can’t confirm external using different subjects, Accept the outcome, don’t validity until additional stimuli and/or situations, reframe your result as studies are completed in field trials or occasions, and “almost significant.” settings. measures. Examine the experimental Researchers can increase and What is a multivariate procedures for design flaws. verify the external validity of design? laboratory findings using If the design appears sound, A multivariate design studies aggregation, multivariate decide whether the multiple DVs. designs, nonreactive hypothesis was reasonable. measurements, field For example, a study of experiments, and naturalistic How should we handle the repetitive strain places a observation. possibility of faulty computer keyboard at procedures? different distances from the What is aggregation? What subject IV) and measures the Check for possible causes of is its logic? effect on three different a non-significant Aggregation is the grouping muscle groups (3 DVs). together and averaging outcome like: 1. confounding
2. extraneous variables that
increase within- subject’s variability
3. weak manipulation of the
IV
4. inconsistent or flawed procedures
5. ceiling and floor effects
6. insufficient power
How should we handle the
possibility of a faulty hypothesis?
If previous studies supported
the hypothesis and ours did not, look for differences in experimental design or sample.
If there was no previous
support and our design and execution were good, we may have to revise or discard our hypothesis.