Blog

Worked analyses with the R script and the output shown.

• The IQR rule: upper and lower fences for outliers

  How to compute Tukey's upper and lower fences by hand and with R's boxplot.stats(), why the two can disagree, and what to do with a value beyond a fence.

  October 22, 2026, Data cleaning and missing data, 10 min read

• Analysing Likert scale data: items and scales

  How to analyse Likert data in a thesis: when a single item needs an ordinal method, when a summed scale can take a t-test, and a worked ordinal regression in R.

  October 22, 2026, Statistics for a thesis or dissertation, 12 min read

• Linear regression assumptions and how to check them

  The four assumptions of linear regression, the R check and the fix for each, and which failures bias the estimates and which only distort the standard errors.

  October 22, 2026, Assumptions and diagnostics, 16 min read

• Logistic regression assumptions, real and invented

  Which assumptions logistic regression really makes, which ones it never made, and how to check each one in R, worked through on the birthwt data.

  October 22, 2026, Assumptions and diagnostics, 13 min read

• One-sample t-test: comparing a mean to a reference

  When a one-sample t-test fits, where the reference value has to come from, and a worked R example on Newcomb's 1882 light measurements with Cohen's d.

  October 22, 2026, Comparing groups, 8 min read

• How to handle outliers before a statistical test

  A decision rule for outliers before a test: correct errors, exclude by a written criterion, keep genuine extremes, with seeded data run with and without them.

  October 22, 2026, Data cleaning and missing data, 13 min read

• How to find a p-value from a test statistic

  Turn a reported t, z, F or chi-square statistic into its p-value by hand and with R's pt(), pnorm(), pf() and pchisq(), compared with full test output.

  October 22, 2026, P-values and inference, 11 min read

• P-values and confidence intervals: report both

  What a p value and a confidence interval each tell you, how they relate, how to read p = 0.06, and how to report both, with a worked example in R.

  October 22, 2026, P-values and inference, 24 min read

• Paired vs unpaired t-test: which design you have

  How to tell whether your two sets of measurements are paired or independent, the designs people misclassify, and what ignoring the pairing costs, shown in R.

  October 22, 2026, Comparing groups, 13 min read

• Parametric test assumptions: checks and fallbacks

  The assumptions behind t-tests, ANOVA and regression, how to check each without the pre-test traps, and what to do when one fails, with a worked Welch ANOVA.

  October 22, 2026, Assumptions and diagnostics, 26 min read