Res & Stats Designs 2

Res & Stats Designs 2

1. If you make specific predictions, can keep critical cut-off for significance at 0.05

   • More comparisons = greater likelihood of getting a significant result by chance
     • Need correction for multiple comparisons
       • To correct for this risk, you can chose a more stringent level of significance (like the Bonferroni), where you divide 0.05 by the number of follow-up comparisons you make (e.g. 0.05/2 = 0.025).
   • If not, then cut off ahs to be more conservative, e.g. 0.01

2. ONE IV, MORE THAN TWO LEVELS - 2 stage analysis

   • FIRST step: Omnibus significance - see if there's a difference overall?
     • Main effects (each IV separately)
     • Interaction (impact of one IV on the other
   • SECOND STEP: FOLLOW UP TESTS to see if there are differences between pairs of groups/conditions
     • Identical to inferential tests
     • To interpret any interaction
     • To interpret the direction of any effects
   • BETWEEN Subs
     • Para metric: One way ANOVA
     • Non-Parametric: Kruskal-Wallis
   • WITHIN SUBS, same people in all conds.
     • Parametric: One way rptd measures ANOVA
       • Non-para: Friedman

3. “Two independent variables interact if the effect of one of the variables differs depending on the level of the other variable.”

4. Mixed design

   • Only poss to analyse with ANOVA
     • Data has to be parametric / meet assumptions of ANOVA

5. Test of cause and effect. Effect of one variable on another

6. MIXED DESIGN

   • As with on IV 2+ levels, you check for main effects first, then for differences between pairs or groups / conditions
   • When there is an interaction, the main effects can look weaker.
     • Follow up tests would show a difference between Group 1 and Group 2, but ONLY in condition 2. Not in condition 1. This shows that the difference between the groups depends on the condition.
   • Mixed
   • More than one IV
   • Some IV's are within, and some are between

7. ONE IV, TWO LEVELS

Imagens e arquivos do mapa