I. Why do Experiments? Experimenter creates differences Control third variable problems--cause and effect II. Independent and Dependent Variables A. Example Hypothesis: Adding fertilizer will speed up tree growth Test: compare with & without fertilizer IV-amount of fertilizer (2 levels: some vs none) DV-tree growth measured as height in mm II. Independent and Dependent Variables B. Levels of IV 1. Choosing levels-levels=ways IV manipulated Fixed levels Continuous IV: Large range, interpolation vs. extrapolation low high low medium 1
II. Independent and Dependent Variables B. Levels of IV 1. Choosing levels-levels=ways IV manipulated no guarantees Interpolation Extrapolation % CR % CR 10 100 Arousal level 10 30 Arousal level II. Independent and Dependent Variables B. Levels of IV 2. Levels vs. more IVs fertilizer administration: none, some-2 levels of 1 IV caffeine dosage- 5 different dosages- 5 levels of 1 IV time of day-(early morning, noon, late afternoon)- 3 levels of 1 IV caffeine dosage X time of day- 2 IVs but 15 testing conditions One IV: conditions=levels More IVs: conditions=combinations of levels 2
Internal Validity -- Truth of causal conclusions in experiment Control- eliminating influence of other variables that could change the results Confounding variable-systematically differs between conditions (could explain differences) 1. Tree - same strain, age, size, history etc. 2. Equating groups in humans-subjects not identical 3. Strategies for equating groups in humans a. Random assignment to groups Sample: Group 1 Group 2 Groups equal on all variables, with large groups (n=10+) 3
3. Strategies for equating groups in humans b. Matched groups-match on chosen variable Sample: 1. Pair them up 2. Randomly assign one from pair to each group. Group 1 Group 2 Groups equal on the variable 3. Strategies for equating groups in humans c. Within-subject designs- same people in all conditions Condition 1: Condition 2: 4
4. Individual differences within a condition Not a confound B. Environmental factors 1. Tree example- same soil, water, humidity, & light same on everything that could influence DV, except IV 2. Characteristics of the testing situation human experiments: physical setting, social setting Confounds-Systematically different between conditions random variations aren t confounds C. Stimulus characteristics 1. Tree Example- characteristics of the fertilizer 2. Equating aspects of stimulus except IV Check the IV manipulations Animal words: aardvark, rhinoceros, platypus, hippopotamus, tyrannosaurus Tool words: hammer, wrench, pliers, pencil, saw Word=equal on length, familiarity, age of acquisition Color on emotion=colors equally bright 5
D. Measurement factors 1. Tree example- measure height visual comparison, ruler, micrometer Same way in all conditions 2. Reliable and valid measures of DV Reliability & Validity same across conditions F. Experimenter factors 1. Tree example- caring for trees same way 2. Treatments carefully scripted- treatments always alike 3. Experimenters blind- Experimenter biasunconsciously create treatment differences or bias measurements 6
Types of Problems so far: Problems with operational definitions: Reliability & Validity You are measuring/manipulating something other than what you intend. Problems with confounds (internal validity): Systematic differences between groups/conditions other than the IV Something else might have caused your results. Problems with samples: Non-representative: Limits generalizability (external validity) Variability within groups: not really a problem Larger effects needed for significance Step 4-Selecting a Method-Specific Exp. Designs A. When to use - no repetition B. True experimental designs 1. Post test only design- 2 or more equivalent groups Each group - one level of IV, measure DV 7
Step 4-Selecting a Method-Specific Exp. Designs B. True experimental designs 1. Post test only design- Example: different cues to reduce tip-of-the-tongue. Repeat question Picture of the object Initial letter of word 11% resolved 15% resolved 47% resolved Good internal validity if: equivalent groups & no confounds Step 4-Selecting a Method-Specific Exp. Designs C. Quasi-experimental designs - Correlational Designs that use groups 1. Nonequivalent control group design 2+ groups, not equivalent selection bias Examples: Schools, Bonding 8
Step 4-Selecting a Method-Specific Exp. Designs C. Quasi-experimental designs - Correlational Designs that use groups 2. Cross sectional design - effects of age or time Subject variables - SV SV = age 15 year-olds --> DV 45 year-olds --> DV 75 year-olds --> DV Cohort effects Step 4-Selecting a Method-Specific Exp. Designs C. Quasi-experimental designs - Correlational Designs that use groups 3. Other subject variables (SVs) gender, personality, intelligence, experience, education level random assignment is not possible 9