Experimental Design • Experiment: A type of research study that tests the idea that one variable causes an effect on another variable. Anatomy of an Experiment Example Memory Cues N1 = 10 M1 = 16.2 S1 = 2.49 No Memory Cues N2 = 10 M2 = 9.9 S2 = 2.33 Independent variable = Memory Training Group Dependent variable = Memory for personal history Anatomy of an Experiment Example Experimental Group Memory Cues N1 = 10 M1 = 16.2 S1 = 2.49 Control Group No Memory Cues N2 = 10 M2 = 9.9 S2 = 2.33 Internal Validity • The study allows the researcher to determine that on variable causes an effect on another variable. Conditions to establish internal validity 1. Time-Order relationship Cause Effect I.V. D.V. Conditions to establish internal validity 2. No alternative explanations • The difference between the means is due only to the independent variable. • Anything else represents a threat to the internal validity of the study Threats to internal validity • Non-equivalent control group – Confound: A way in which the groups differ from each other, other than the independent variable. – Controlling for confounds • 1. Random assignment to groups • 2. Matching Threats to internal validity • Floor or Ceiling effects – The independent variable has made the groups different from each other, but the dependent variable is unable to detect it. – Floor effect: The test is so difficult that everyone gets a very low score. – Ceiling effect: The test is so easy that everyone gets a high score. – They make the means closer together than they should be. Threats to internal validity • Experimenter effect – The experimenter gives an indication of what they want or expect the subject to do in a particular condition. • Participant effect – The participant changes their behavior to fit what they think the researcher is studying. Ways to address experimenter and participant effects – Single-blind study: The participant doesn’t know which condition they’re in. • Example: a placebo-controlled condition. – Double-blind design: Neither the participants or the researcher knows which condition the subject is in. External Validity • The results of the study are generalizable 1. Generalization to different samples – Get the same results if repeat the same study with a different sample (from the same population) – Replication External Validity 2. Generalization to different populations – Get the same results if repeat the same study with a sample from a different population 3. Generalization to different settings – Get the same results under different conditions – The effect is observed in more than one setting – Example: The effect is observed in real life, not just in the laboratory Independent Samples T-Test • Tests the difference between two sample means Memory Cues N1 = 10 M1 = 16.2 S1 = 2.49 No Memory Cues N2 = 10 M2 = 9.9 S2 = 2.33 Prediction of the researcher: The mean of the Memory Cues Group will be significantly higher than the mean of the No Memory Cues Group. Independent Samples T-Test Prediction of the researcher: The mean of the Memory Cues Group will be significantly higher than the mean of the No Memory Cues Group. – Example of a one-tailed test – One-tailed test: One mean is predicted to be higher or lower than the other one. – Two-tailed test: One mean is predicted to be different from the other one. Independent Samples T-Test Prediction of the researcher: The mean of the Memory Cues Group will be significantly higher than the mean of the No Memory Cues Group. – Example of a one-tailed test – Alternative hypothesis: The mean of the Memory Cues Group is significantly higher than the mean of the No Memory Cues Group. – Null hypothesis: The mean of the Memory Cues Group is not significantly higher than the mean of the No Memory Cues Group. Independent Samples T-Test – No way to know for sure which hypothesis is true. – We can know the odds that the null hypothesis is true. – We can decide how unlikely the null hypothesis would have to be before we can’t believe it anymore. That’s the Alpha Level of the test. – “α = .05” means “Reject the null hypothesis if the odds are less than 5% that it’s true” Independent Samples T-Test An independent samples t-test tells you if the odds are less than 5% that the null hypothesis is true. 1. Find the number we’re making our decision about • It’s the difference between the two group means • M1 – M2 = 16.2 – 9.9 = +6.3 • We’re comparing this number to a difference of zero. 2. Convert that number to a standard score – In SPSS, t = +5.85 – The difference between the two sample means is 5.85 standard deviations above a difference of zero. Independent Samples T-Test 3. Find how far from zero that number needs to be to be significant Critical Value for t • We predicted that this difference would be in the positive direction, so it’s a one-tailed test. • α = .05 • Degrees of freedom = N1 + N2 – 2 10 + 10 – 2 = 18 • Critical value = +1.73 • Decision rule: If t ≥ +1.73, reject the null hypothesis. Independent Samples T-Test Conclusion: The mean of the Memory Cues Group is significantly higher than the mean of the No Memory Cues Group, t (18) = 5.85, p < .05.