Description
Conventions for writing results: means with SDs, named test statistic with degrees of freedom, exact p against a stated alpha, effect size, confidence intervals and consistent precision.
Results are only useful if readers can check them. The APA conventions require, for every comparison, the descriptive statistics (means with SDs, or medians with ranges for non-parametric data), the test statistic with its degrees of freedom, the exact p value judged against an alpha stated once, and an effect size; point estimates come with 95% confidence intervals. Smyth adds practical rules: report assumption tests only if they changed the analysis, use 'significant' only in the statistical sense and never 'highly significant', round inferential statistics to two decimals without false precision, drop the leading zero for values that cannot exceed one, put complex analyses in tables, never show the same data in a figure and a table, and report effect size and power for non-significant results. The conventions make UX quantitative findings comparable and auditable.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Siloed
- Goal
- Communicate findings
- Stage
- Evaluation, Post-launch
- Data
- Quantitative
- Effort
- Hours
Steps
- State the alpha level used for all tests at the start of the results.
- For each comparison give group means and SDs, then the test statistic with degrees of freedom, exact p, tails if one-tailed, and an effect size, for example t(16) = 2.41, p = .028, d = 1.14.
- For ANOVA give F(df1, df2), p, an effect size such as omega-squared or partial eta-squared, and MSE; report interactions before main effects.
- For chi-square give chi-square(df, N = n) and p; for correlations give r with its confidence interval and p.
- Give point estimates with 95% confidence intervals and round inferential statistics to two decimals, percentages to whole numbers.
- Put complex analyses in a table, keep simple results in text, and never duplicate the same data in a table and a figure.
- For non-significant results report the effect size and its interval and consider power before concluding there is no effect.
Sources
- The Principles of Writing in Psychology (Smyth, 2004)
Also known as
- Reporting statistics in APA style
- APA statistical reporting format
- Results section statistics conventions
Categories
- Analytics & Behaviour
- Statistics & Modelling