Plotwright scientific guidesEvidence-backed statistics and scientific-graphing examples for researchers.https://plotwright.furiga.com/feed.xml2026-08-08T17:10:00+02:00Furiga GmbHhttps://plotwright.furiga.com/Michaelis–Menten: fit Km and Vmax without linearizinghttps://plotwright.furiga.com/guides/michaelis-menten-km-vmax.html2026-08-08T17:10:00+02:002026-08-08T17:10:00+02:00Six substrate concentrations fitted directly and via the double-reciprocal plot, showing a 12% shift in Vmax and 23% in Km despite r² = 0.998.Before-and-after plots: draw the pairing, do not average it awayhttps://plotwright.furiga.com/guides/before-after-paired-plot.html2026-08-08T17:10:00+02:002026-08-08T17:10:00+02:00Eight subjects measured twice, all eight decreasing: the paired analysis gives P = 0.0000254 while ignoring the pairing gives P = 0.0991 on the same sixteen numbers.Standard curves: fit the right range, then interpolatehttps://plotwright.furiga.com/guides/standard-curve-interpolate-unknowns.html2026-08-08T16:40:00+02:002026-08-08T16:40:00+02:00Seven visible calibration standards fitted twice: r² = 0.926 across the full 0–64 ng/mL range hides systematic curvature that the residuals expose, while the 0–16 linear range gives r² = 0.997 and a slope almost twice as steep.Grouped bar graphs: which bar, which errorhttps://plotwright.furiga.com/guides/grouped-bar-graph-error-bars.html2026-08-08T16:35:00+02:002026-08-08T16:35:00+02:00A 2 × 2 design with three replicates per cell, with all twelve values visible, showing what SD, SEM and 95% confidence-interval bars each claim — and why the smallest bar is the least honest at n = 3.Shapiro–Wilk and Q–Q plots: P > 0.05 is not proofhttps://plotwright.furiga.com/guides/shapiro-wilk-normality-test-qq-plot.html2026-08-02T19:28:00+02:002026-08-02T19:28:00+02:00Use one fixed mildly skewed sequence to see Shapiro–Wilk move from P = 0.929 at n = 12 to P = 0.00690 at n = 48, then interpret the Q–Q plot and model quantity without treating a threshold as proof.Paired t test vs Wilcoxon signed-rank: do not swap the question silentlyhttps://plotwright.furiga.com/guides/paired-t-test-vs-wilcoxon-signed-rank.html2026-08-02T19:05:00+02:002026-08-02T19:05:00+02:00Use Student's ten matched patients to compare mean-difference and signed-rank inference, reproduce exact results, and see why a normality pre-test should not silently change the scientific question.Technical vs biological replicates: why n is 3, not 12https://plotwright.furiga.com/guides/technical-vs-biological-replicates.html2026-08-02T15:24:00+02:002026-08-02T15:24:00+02:00Use 24 visible assay readings from six simulated cultures to see why flattening technical replicates produces P = 0.00066 while the nested analysis reports P = 0.162.Repeated-measures ANOVA vs mixed effects: what changes with one missing value?https://plotwright.furiga.com/guides/repeated-measures-anova-vs-mixed-effects.html2026-08-02T15:05:00+02:002026-08-02T15:05:00+02:00Use 39 observed weights from eight matched chicks to see why complete-case repeated-measures ANOVA drops an entire subject, what a random-intercept mixed model retains, and which missingness and covariance assumptions remain.ANOVA vs Kruskal–Wallis: means or ranks?https://plotwright.furiga.com/guides/anova-vs-kruskal-wallis.html2026-08-02T14:21:00+02:002026-08-02T14:21:00+02:00Use all 30 PlantGrowth observations to compare the mean-based one-way ANOVA with rank-based Kruskal–Wallis inference, including the different null hypotheses, assumptions, exact outputs and editable projects.Tukey vs Dunnett: every pair or each treatment versus control?https://plotwright.furiga.com/guides/tukey-vs-dunnett-test.html2026-08-02T13:50:00+02:002026-08-02T13:50:00+02:00Use all 30 PlantGrowth observations to compare the all-pairs Tukey family with control-focused Dunnett inference, including exact simultaneous confidence intervals, adjusted P values and two editable projects.Bland–Altman vs correlation: does high r mean agreement?https://plotwright.furiga.com/guides/bland-altman-vs-correlation.html2026-08-01T19:46:00+02:002026-08-01T19:46:00+02:00Use Bland and Altman's original 17-subject PEFR data to see why r = 0.943 does not establish agreement, reconstruct bias and 95% limits of agreement, and separate statistical estimates from pre-specified practical tolerances.SD vs SEM error bars: spread is not precisionhttps://plotwright.furiga.com/guides/sd-vs-sem-error-bars.html2026-08-01T19:20:00+02:002026-08-01T19:20:00+02:00Use 25 transparent values to distinguish observed spread from mean precision, compare SD, SEM and t-based 95% confidence intervals, and avoid pseudoreplication hidden behind narrow error bars.Welch vs Student's t test: do not let a variance pre-test choosehttps://plotwright.furiga.com/guides/welch-vs-student-t-test.html2026-08-01T14:01:00+02:002026-08-01T14:01:00+02:00Use 24 transparent synthetic values to see why Welch and pooled Student t tests can disagree when variance and sample size are both unequal—and why a preliminary variance test should not choose between them.Fisher exact vs chi-square: use expected countshttps://plotwright.furiga.com/guides/fisher-exact-vs-chi-square.html2026-08-01T13:45:00+02:002026-08-01T13:45:00+02:00Use R's eight-cup tea-tasting table to distinguish observed from expected counts, compare exact and asymptotic inference, and avoid applying independent-table tests to paired designs.Pearson vs Spearman: linear or monotonic correlation?https://plotwright.furiga.com/guides/pearson-vs-spearman-correlation.html2026-08-01T13:24:00+02:002026-08-01T13:24:00+02:00Use R's nine-lot tuna-quality example to distinguish linear Pearson correlation from monotonic Spearman rank correlation, inspect every pair, and report small-sample inference conventions honestly.Mann–Whitney U test: not automatically a median testhttps://plotwright.furiga.com/guides/mann-whitney-u-test-median.html2026-08-01T09:22:00+02:002026-08-01T09:22:00+02:00Use the canonical chorioamnion example to see what U counts, distinguish stochastic ordering from a median difference, and report exact inference, probability of superiority and a qualified Hodges–Lehmann shift.One-way ANOVA and Tukey: read the F test correctlyhttps://plotwright.furiga.com/guides/one-way-anova-tukey-plantgrowth.html2026-08-01T08:56:00+02:002026-08-01T08:56:00+02:00Use all 30 PlantGrowth observations to separate the omnibus one-way ANOVA from Tukey-adjusted pairwise differences, simultaneous confidence intervals and design assumptions.Two-way ANOVA interaction: interpret it firsthttps://plotwright.furiga.com/guides/two-way-anova-interaction.html2026-08-01T08:44:00+02:002026-08-01T08:44:00+02:00Use all 60 ToothGrowth observations to interpret a supplement-by-dose interaction, distinguish cell means from collapsed main effects, and plan defensible follow-up comparisons.ROC curve and AUC: choosing a diagnostic cutoffhttps://plotwright.furiga.com/guides/roc-curve-auc-threshold.html2026-08-01T07:02:00+02:002026-08-01T07:15:00+02:00Use the canonical Hanley–McNeil 109-image example to connect ROC AUC with ranking, examine sensitivity and specificity at every rating threshold, and separate a statistical cutoff from clinical policy.Kaplan-Meier survival curves: censoring and log-rankhttps://plotwright.furiga.com/guides/kaplan-meier-survival-curve.html2026-08-01T06:51:00+02:002026-08-01T07:15:00+02:00Use the canonical 6-MP remission dataset to distinguish events from censoring, read a risk table, interpret median survival and understand the limits of a log-rank P value.How to calculate IC50 and EC50 from a dose-response curvehttps://plotwright.furiga.com/guides/ic50-ec50-dose-response.html2026-08-01T06:35:00+02:002026-08-01T07:15:00+02:00Fit a seven-concentration dose-response example with a variable-slope four-parameter logistic model, then check its range, plateaus, replication, diagnostics and profile-likelihood confidence interval.Paired vs unpaired t test: the same data, two answershttps://plotwright.furiga.com/guides/paired-vs-unpaired-t-test.html2026-07-31T21:40:00+02:002026-08-01T07:15:00+02:00The same canonical sleep data produce P = 0.002833 when the patient pairing is preserved and P = 0.07919 when it is discarded. Work through the design, exact result, graph and limitations.Anscombe's quartet: why you must graph your datahttps://plotwright.furiga.com/guides/anscombe-quartet.html2026-07-31T21:40:00+02:002026-08-01T07:25:00+02:00Four datasets share nearly identical regression summaries while their graphs reveal a line, a curve, an outlier and a high-leverage point.