Flashcards
Stats revision
Cards in this deck (55)
- 1Univariate ANOVAsUsed to reduce error rates and avoid false positives (type 1s); we can see if there is an overall difference
- 2How do we avoid a type 1 error?Correct the P value (normally through Bonferroni), but could increase the possibility of false negatives
- 3Between-Subjects ANOVAUsed when need to explore 3+ independent conditions on a single variable; tells us overall condition, not what conditions differ from each other
- 4What do you need to run a Between-Subjects ANOVA?Normally distributed data (interval/ratio data) and homogeneity of Variance
- 5Do we want Levene's to be sig or non-sig?Non-significant, it means we have H of V
- 6What effect does a small sample have on the effect size?Effect size will be bigger
- 7How do we correct this?Run a Bonferroni
- 8What graphs would you use to represent unrelated groups and which for mixed designs?Simple bar chart for unrelated, clustered for mixed
- 9What does it mean if the error bars are tight?Low variance
- 10What is standard error?Measure of spread
- 11What does it mean if 2 separate bars don't don't overlap?There is a significant difference
- 12What is Cohen's D used for?Used to compare the magnitude of differences between 2 means
- 13If you have a Cohen's D of 0.2+ but less than 0.5 what effect size do we have?Small Effect
- 14If you have a Cohen's D of 0.5+ but less than 0.8 what effect size do we have?Medium Effect
- 15If you have a Cohen's D of 0.8+ what effect size do we have?Large Effect (can be as much as 1)
- 16What are Pearson's Correlation Coeffients?0.1 - 0.3 = small
0.3 - 0.5 = medium
0.5 + = large - 17When is a One Way Within Subjects ANOVA used?When there is a single within subjects variable with 3+ levels
- 18What does a One Way Within test?Whether differences in scores between the within subjects conditions are higher than differences between scores due to chance factors and controlling for ind diffs
- 19What would you do instead if you only had 2 levels?Paired Samples t-test
- 20What would you do instead if the data was non-parametric?Friedman test
- 21What assumptions do you need to meet before you run the ANOVA?Normally distributed data (ratio or interval data) and sphericity
- 22What is sphericity?Equality in differences of variances between IV levels
- 23Would we have sphericity if you only had 2 levels?Yes, sphericity is automatically assumed
- 24Do we want Mauchley's test to be significant or non-significant?Non-significant
- 25What is a MANOVA?Multivariate Analysis Of VAriance
- 26When would we use a MANOVA?When you have 1 IV and many DVs
- 27What's the problem with having many DVs?You increase the probability of having a type 1 error
- 28How do you reduce the possibility of having a type 1 error?Only use 1 test
- 29What does a MANOVA assessIf an IV has an overall effect on variables
- 30How does a MANOVA assess this?Using a single P value to see which specific aspects affect the result
- 31What assumptions does a MANOVA have?The usual ANOVA assumptions (normally distributed data, interval or ratio, H of V)
- 32Can a MANOVA be within subjects?No, its always between
- 33What is Box's test?Equality in covariance matrices- only presents a problem when it is <.001
- 34If we have 2 levels of an IV do we need to do post-hocs?No
- 35What do we use if 1 variable does not show equality of variance?Use Tanhane's T2
- 36What is an ANCOVA?ANalysis of CO-VAriance
- 37What are covariates?Variables that associate with the DV but are not part of the experimental manipulation
- 38Why could they be an issue?We don't want them to interact with the IV
- 39Why are co-variates used?To reduce within group error variance (more likely to find a significant effect of manipulation and eliminate confounds)
- 40What assumptions do we have?Usual ANOVA assumptions (Normally distributed data, interval or ratio data, H of V); independence of covariate and IV effect; independence of covariate and homogeneity of regression slopes
- 41How do we get make sure we have homogeneity of regression slopes?We run an ANOVA on the covariate and want there to be a non-significant result
- 42What is homogeneity of regression slopes?The relationship between the covariate and DV is constant across all groups
- 43What happens with a Complex Between subjects design?If there is more than 1 IV, then we can expect there to be an interaction
- 44How can you see if there is an interaction when the data is presented graphically?Lines converge, diverge or cross
- 45If you have a large sample, would you have greater or smaller power?Greater
- 46What assumptions need to be met?Normally distributed data; interval/ratio data; H of V
- 47What will happen in SPSS if you get a significant interaction?It won't compute post hoc tests for pairwise comparisons within the interaction
- 48When do we use a Complex Within Subjects ANOVA?When we have more than 1 IV
- 49Do all participants have to contribute a score in every condition?Yeah
- 50What assumptions need to be met?Normally distributed data; ratio/interval data; sphericity (which is checked for main effects and interactions)
- 51When do we need sphericity?When we have 3+ levels
- 52How do we conduct post hocs?Use the syntax method within SPSS or use a series of T-Tests
- 53When do we use mixed ANOVAs?When we have within and between subjects IVs
- 54What assumptions need to be met?Usual ANOVA assumptions (normally distributed data; interval or ratio data; H of V; sphericity; equality of covariance matrices)
- 55How would you conduct post-hocs?Paired samples T-tests, conducting them separately for each condition