Mapa mental
stats week 4
Contenido del mapa
stats week 4
RM ANOVA
- if you did a one way RM ANOVA with 2 levels, you'd obtain the same result as paired t-test F = t2
- extension of paired t-test
- assumptions - normality - the distribution of difference scores under each IV level should be approximately normal (unlikely to be a problem when N>30) - sphericity - (homogeneity of covariance) - the variance in difference scores under each IV level pair should be reasonably equivalent (Maulchy's test assess, greenhouse geyser corrects for) - sample size equivalent
- Mauchly's - there is no difference between the covariances under each IV level pair
- what contributes to variance? between IV levels - manipulation of IV, experimental error (random and constant) - within = experimental error (random)
- we subtract the variance due to individual differences from the variance within IV levels
- applied when theres 3 or more IV levels as long as only one IV
Variance between IV levels / variance within IV levels
- total variance is the sum of variance between IV levels (explained by our model) + variance within iV levels (individual differences + error/residual variance)
- F = MSm / MSr - value close to 0 = small variance between relative to within - further from 0 = large variance between relative to within
- total variance is the sum of variance between IV levels (explained by our model) + variance within iV levels (individual differences + error/residual variance)
DF
- the difference between the number of measurements made and the number of parameters estimated - need to calculate for between and within - DFm (between) = k - 1, DFr = DFm x (n-1)
Post hoc
- used to assess which IV level means differ - should only be used when F-value is singificant - a number of tests are available but we use Boferroni
- Partial n2 = SSm / SSm + SSr
- Cohens d = magnitude of the difference between IV level means, expressed in SD units
- Partial n2 = SSm / SSm + SSr
- used to assess which IV level means differ - should only be used when F-value is singificant - a number of tests are available but we use Boferroni
advantages
- needs fewer participants to gain same number of measurements - error variance is reduced (bc we remove individual differences and this variance is elimanted from our model variance) - more power with same number of participants (easier to find significant difference (avoid type 2 error) because with the same variance due to IV manipulation, the resulting F/t value is larger
disadvantages
- order effects = practice effects, fatigue effects, sensitisation, carry-over effects - can control this with counterbalancing - but isn't always possible - extensive pre-study practice, design short experiments, allow time between exposure to IV levels, include control group
Información
- Descripción
- Degree Statistics Mapa Mental sobre stats week 4, creado por Abbey Wrigley el 14/01/2019.
- Creado por
- Abbey Wrigley
- Visitas
- 1
- Valoración
Más sobre Repeated Measures ANOVA