Behavioral Research Chapter 10 Complex Experimental Designs Simple designs • Composed of one indep var that is manipulated with two levels and one dep var which is measured. – Example: IV: Stress vs. no stress (control) • Both measured by a test of cognitive function • Hypothesis: The affects of stress impair cognitive function stress as well as cognitive function would have to be operationally defined as to what was used as a stressor (IV) and what measurement did one use to measure cognitive function (DV) Factorial design • Designs with more than one indep var or factor . – all levels of each indep var are combined with all levels of the other indep var – The simplest type of factorial design is a 2 X 2—has two indep var, each having two levels. Example: 2 X 2 – Indep var 1: difficulty of the task—easy or hard – Indep var 2: attitude of the confederate— helpful or mocking – Dep var: performance on a cognitive task Four Experimental Conditions for 2 X 2 Factorial Design – – – – Easy task – helpful confederate Easy task – mocking confederate Hard task – helpful confederate Hard task – mocking confederate Interpretation of Factorial Designs • Main Effect The impact of each IV on the DV. -The number of main effects depends on the number of Independent Variables. • InteractionThe effect of any combination of two or more IV. on the D.V.. – The effect that an independent variable has on the dependent variable depends on the level of the other independent variable. Example: Factorial design Examining the after-effects of exposure to an irritating noise on several behavioral measures as a measure of frustration: Two levels of each independent variable • Hypothesis: IV one: Irritating noise: loud vs. soft IV two: Predictability: predictable vs. non predicable DV: Number of attempts at difficult puzzle during different noise levels Noise intensity vs. Predictability Loud Soft Group 1 Group 2 Unpredictable Group 3 Group 4 Predictable Noise intensity vs. Predictability Loud Soft 7 8 Unpredictable 3 5 Predictable Calculating Main Effects: Comparing Row and Column Means Column Means: Loud= 5 Soft = 6.5 Row Means: Predictable = 7.5 Unpredictable = 4 Interpretation of Main Effects – A reliable difference in the column means would indicate an effect of noise intensity, independent of noise predictability – A reliable difference in the row means would indicate an effect of noise predictability, independent of noise intensity Interactions • Number of attempts to solve the difficult puzzle was greater when the noise was soft than when it was loud. • However, this relationship was dependent on whether the noise was unpredictable Effects of Predictable vs. Unpredictable Loud vs. Soft Noise on Puzzle Attempts Number of puzzle attempts 10 8 6 Predictable 4 Unpredictable 2 0 Loud VS. Soft Factorial Designs With Manipulated and Nonmanipulated Variables: IV X PV Designs – allow researchers to investigate how different types of individuals respond to the same manipulated variable – E.g., of Participant variables – gender, age, ethnic group, personality characteristics – The simplest IV X PV design includes one manipulated independent variable with at least two levels and one Participant variable with at least two levels – E.g., Participant variable – two different age groups; or males vs. females IV X PV design, Furnham, Gunter, Peterson (1994) – Showed that the ability to study with a distracting task in the room is affected by whether you are more extraverted or introverted – Manipulated var—distraction – Subject var—extroversion or introversion – Measured var—reading comprehension – A repeated measures design was used college students read material in silence and within hearing range of a TV program Results – Overall, students had higher comprehension scores when they studied in silence – Interaction between extraversion and distraction – Without a distraction, the performance of extraverts and introverts was the same – However, extraverts performed better than introverts when the TV was on. Further Considerations in Factorial Designs • If you were to have a 2 x 2 x 2 factorial design, you could look at it as two 2 x 2 designs. – E.g., 2 (instruction method: lecture or discussion) x 2 (class size: 10 or 40) x 2 (gender) – Divide 2 x 2s by gender—2x2 for males and 2x2 for females – Could then look at the main effects and interactions within each of these 2 x 2s )(three main effects) – gender – lecture vs. discussion – class size (small=10; large= 40) Interactions in a 2 X 2 X 2 – Could also look at the interaction in the 2 x 2 x 2 design—have the possibility of 3 simple interactions – Instruction method and class size – Instruction method and gender – Class size and gender » Could also have a three-way interaction, where the effect of the interaction b/t two of the variables differs depending on the particular level of the third variable » Three-way interactions are complicated and hard to interpret F-Statistic • Used in Factorial Designs • Is an extension of the t-test. • It is an analysis of variance that is a more general procedure than the t-test. • When a study has only one independent variable and only two groups using an F or a t makes no difference. • However analysis of variance (ANOVA) is conducted when there are more than two levels of an independent variable and when a factorial design with two or more independent variables is used. • Therefore, the f-test is appropriate for the simplest experimental design as well as more complex. • T-test demonstrates the relationship between two groups and the within group variability • F test is the ratio of two types of variance: – Sytematic variance: deviation of the group means from the grand man which is the mean score of all individuals in all groups. (Grand mean-5.75: Loud=5, Soft = 6.5) • Is small when the differences between group means is small and increases as the group mean differences increase – Error variance: the deviation of the individual scores in each group from their respective group mean. F-Significance • Ratio of Systematic variance over Error Variance. • Therefore you want systematic variance (difference between groups as shown by comparing grand mean to group means) to be high. • Error variance to be low (comparison of individual scores against the group mean) • Low error variance indicates homogeneity within your groups which will increase your F statistic and be more likely to show significant results.