Reporting Statistical Analyses

Graph your data prior to analysis

Draw clear, concise graphs

• Always report d.f. for parametric tests
• Report sample sizes with non-parametric tests
• Degrees of freedom say how much power there is to find a real effect

Report essential info from R, don't copy and paste

p-value is never 0

Saying results are just signficant means nothing - state the facts

Things are never nearly significant

Use sensible decimal places - tell us about the level of accuracy

• For any set of numbers measured in the same way, use same number of decimal places

If there are no signifcant results, don't use a line of best fit

Read table from bottom up - look at interaction results first

There are alwys 2 sets of d.f. w/ ANOVA

Always visually inspect assumptions of ANOVA after analysis:

• Homoegeneity of variances - use residuals vs fitted graph and/or Bartlett test
• Skewed data - use Normal Q-Q and/or Shapiro-Wilk

Convert interaction plots into concide graphs

Conducting several ANOVA is only useful when each is testing a different response

Conditions for a two way ANOVA from a graph:

• Error bars = replicate measures at each level of each factor
• x-axis shows two factors
• Factors are fully crossed - able to test interactions

With repeated measures, make sure you declare repeated measures as a cause of variance