Help shape GoConqr. What should we add next?

Mind map

Stats revision

Marie Lush English

Map content

Stats revision

  1. Is P<0.05?

    • Yes
      • Data is not normal
    • No
      • Data is normal
        • Bartlett's test for homogeneity
          • Is P<0.05?
            • Yes
              • Heteroscedastic
            • No
              • Homoscedastic
        • USE PARAMETRIC APPROACHES
          • ANOVA
            • One way
              • Measures the variance of each group
              • Tests if variance is significantly different from total variance
              • MSb/MSw = Fstat
              • If Fstat > Fcrit, reject H0 meaning there is a signifdif in means of groups
            • Nested
              • Use when there are subgroups
              • Tests for significant variation in means among groups and sub groups
              • EXTENSION OF ONE WAY ANOVA
            • Welch's
              • Alternative to One Way ANOVA
              • Use when data:
                • Is balanced in terms of experimental design
                • Has small number of obs in each group
                • Shows too much hetroscedacity
            • Two way
              • Tests 3 H0's at the same time
              • Look at interaction P-value first - if less than 0.05 there is an interaction between the two nominal variables
          • T-Tests
            • Two Sample
              • Must pass Bartlett's
              • Test if means of two groups are statistically different
              • One way ANOVA post hoc
                • ANOVA says there is a difference, T-Test shows where the difference is
              • DF = n-2
            • Paired
              • Pairs of data - eg. before and after
              • Test is mean difference between pairs is 0
              • If Tstat > Tcrit, then P<0.05 so reject H0
              • Null - mean difference is 0
            • One Sample
              • Tests if mean of data is significantly different to the expected value
              • H0 = no difference in data from experimental value
              • If Tstat > Tcrit, reject H0 and assume mean of data is significantly different to expected value

Information

Description
Mind Map on Stats revision, created by Marie Lush on 01/01/2020.
Created by
Marie Lush
Views
0
Rating

More on Factorial ANOVA and Interactions

Feedback

What is it about?

What should we add next?