Week 9-Complex Between Subjects Flashcards
Can you have more than 2 IVs in a complex design?
In theory yes. But expanding the number of IVs makes interpretation of interaction effects very challenging. (i.e., more than 2 IV makes study lack clarity overall).
What is Between Subjects ANOVA?
-Complex research designs- more than one IV
-This is because we expect an interaction
-I.e. The effects of one IV is dependent upon the other IV
-Under condition A of IV one the effect of IV two will be different than when it is under condition B.
-Interaction between 2 independent variables working together to produce an effect on the dependent variable
Define main effects and interactions
Main effects: The effect of each IV in isolation (regardless of the other IV) - You report 2
The interaction: effects of the two IV’s working together - You report 1
What is the importance of the sample size?
-Sample size: This will effect the power of your study, greater numbers = greater power
-If you are testing lots of IV’s with lots of levels you need more power
-You need more power if you expect main effects AND interactions (rather than just an interaction)
-Rule of thumb – 30 participants per cell to detect a moderate to large effect size
-Level example could be 3x3 design=9 cells=270 participants
-The sample size issue is greater when we use between subjects ANOVA than it is in a within subjects design
-This is because everybody contributes just one score/figure to the analysis
-We can think about this in terms of cells
-If there is any way to make part of study repeated measures, it reduces the number of participants needed
When should descriptive statistics be reported in text? – Simple ANOVAs
-For simple ANOVAs, always report the descriptive statistics (means and SD) in text when breaking down a significant effect of the IV on the DV (only when it is significant).
-The exception to reporting means and SDs in text is if in the exam, you are asked to create a descriptive statistics table which includes the means and SDs of the data. In this case, descriptive statistics which are stated in the table, do not need to be repeated again in text, in response to another exam question. This is true for both simple and complex ANOVA analyses.
When should descriptive statistics be reported in text? – Complex ANOVAs
For complex ANOVAs, SPSS can be funny with the descriptive statistics that it provides.
-When the variable is a between-subjects variable, SPSS will provide the mean and standard deviation for all conditions of the variables and a ‘total’ mean and standard deviation for each variable.
-This means that we can report the means and standard deviations for each condition when breaking down main effects.
-When the variable is a within-subjects variable, means and standard deviations are presented for each condition, but we don’t get a ‘total’ mean and standard deviation for each variable.
-This means that we cannot report the means and standard deviations for each condition when breaking down main effects.
-However, we can view the estimated marginal means output which gives us the mean for each condition – this can be used to interpret the direction of the effect. (no SD with it)
When should descriptive statistics be reported in text? – Complex ANOVAs Interactions
-For the interaction effect, the overall interaction effect should not have any means of standard deviations reported with it (as there are no means and SDs which make sense to report with the overall interaction effect).
-For post-hoc testing of the interaction effect, however, means and SDs should be given, when presented in the output.
-For paired-samples t-tests, we are given the means and SD for each comparison in the ‘paired samples statistics’ box, so these should be reported in text, when reporting the t-tests
-However, when breaking down the interaction effect using estimated marginal means (like I’ve shown in this lecture), the SDs of each comparison are not given.
-For complex between subject ANOVA, report the means and SDs from the descriptive statistics table.
Give a summary of this topic
-Complex designs involve two or more independent variables
-Complex between subjects designs tend to require a lot of participants
-When analysing complex designs we report main effects- the influence of each IV on the DV in isolation
-The interaction, the influence of the two IV’s in combination on the DV
-Post hoc tests are required to explore a SIGNIFICANT main effect BUT ONLY IF IT HAS THREE OR MORE LEVELS
-Post hoc tests are always required to explore a SIGNIFICANT interaction (To do this requires syntax editing)
