More complete statistics, more consistent axes and figure details
PivoGraph 1.7 adds one-way ANOVA and descriptive statistics to Column tables, and the t-test gains Mann-Whitney and Wilcoxon nonparametric options and effect sizes. XY tables gain area under the curve and correlation analyses, and the survival curve comparison adds a test for trend, 95% CIs for median survival, and the ratio of median survival. Continuous axes gain minor ticks, and survival curves get new default axes, error bars, and adjustable censor ticks. Spikes, area fills, before-after graphs, and floating bars have more complete appearance options, and pie charts and other parts-of-whole graphs are drawn at the size you set. The website adds a Learn section, and the validation page now compares results with R value by value. The project file format moves to v3, and older files migrate automatically when opened.
New features
One-way ANOVA
Analyses on Column tables include “One-way ANOVA (three or more groups)”. Based on the experimental design (independent groups or the same subject in each row), the distribution of the data, and whether the groups have equal SDs, it runs ordinary ANOVA, Welch and Brown-Forsythe ANOVA, repeated-measures ANOVA with the Geisser-Greenhouse correction, the Kruskal-Wallis test, or the Friedman test; nonparametric tests on small samples report exact P values. Multiple comparisons can cover all pairs, comparisons with a control, or selected pairs, with a recommended correction for each case (Tukey, Dunnett, Šídák, Dunnett T3, Dunn, and others, plus Holm-Šídák, FDR, and no correction). Results include tests for equal variances, the ANOVA table, a summary of each group, the comparison table, and significance letters. “Add brackets to graph” places only the significant comparisons by default and stacks them by span; when there are many levels, the graph title moves up to make room. Brackets you draw yourself also take their stars from the corrected results.
Descriptive statistics
Column tables gain “Descriptive statistics”, listing N, mean, SD, SEM, the confidence interval of the mean, median, and quartiles for each column, with CV, geometric mean, the confidence interval of the median, skewness, kurtosis, and more when expanded.
Nonparametric t-tests and effect sizes
“t-test (two groups)” gains a Distribution option. With the nonparametric choice, unpaired data use the Mann-Whitney test and paired data use the Wilcoxon matched-pairs signed rank test; when each group has fewer than 50 values, the P value is exact, including when there are ties. Results include rank sums, U, the median of each group, and the Hodges-Lehmann difference with its confidence interval, or the sums of positive and negative ranks, the median difference, and the Spearman correlation of the pairs. Results of the normal t-test add R², Cohen's d and Hedges' g (dz for paired data), an F test comparing the two variances, and the correlation of the pairs. Stars on significance brackets follow the selected test.
Area under the curve and correlation on XY tables
XY tables gain “Area under curve”, calculated for each data set with the trapezoid rule against an adjustable baseline Y, reporting the area above and below the baseline and the net area; when every X has replicates, the standard error of the net area is also given. “Correlation (X vs. Y)” calculates Pearson r or Spearman ρ between X and the Y of each data set, with R², a two-tailed P value, and a 95% confidence interval for Pearson r.
Survival comparison: test for trend, median CIs, and ratio of medians
With three or more curves, “Compare survival curves” adds the log-rank test for trend, which scores the groups in table order and suits ordered groups such as increasing doses. Each median survival time comes with its 95% CI, and an upper limit the curve never reaches is shown as ∞. With two curves, the ratio of median survival and its 95% CI are added. With “Add results text to graph”, the P value lines include the test for trend, and the median survival line includes the 95% CI.
Censor ticks on survival curves
In “Staircase with ticks”, the short vertical marks at censored observations become a separate “Censor ticks” card with color, line width, length, and direction (up, down, or both; up by default). The color follows the data set's border color by default, and the legend shows both the line and the tick. This graph type now draws no error bars by default; SE or 95% CI remain available in Format Plot.
Minor ticks and end gaps on axes
The continuous X axis and the Y axes in Format Axes gain “Minor ticks”: automatic, none, or 2, 4, 5, or 10 per interval. The part of a minor tick outside the axis line is half that of a major tick at every zoom level and export resolution, and minor ticks have no labels. The continuous X axis also gains “Leave a gap at both ends” (off by default), which extends the automatic range by about half an interval at each end so the Y axis no longer covers the start of a curve at time 0.
More appearance options for spikes and area fills
Spike graphs can set the width (pt, default 4, no longer tied to the X spacing), border color, and border thickness in Format Graph. Their baseline uses the same “Automatic / Spikes begin at Y= / Hide baseline” setting in Format Plot as bar graphs, and a manual baseline draws a line across the plot. Both spike graph types can add symbols and connecting lines (off by default; the legend key changes accordingly). Area fills gain a Position setting: below curve (to the lower end of the Y axis), above curve (to the upper end), and to the X axis (to Y = 0).
Mean bars on before-after graphs and points on floating bars
“Before-after” gains a bar card, off by default and available for each data set, that draws a semi-transparent bar at each column's mean behind the lines and symbols. “Floating bars (min to max)” gains a “Symbols” card, off by default, that overlays the individual values on the floating bars; the bars become semi-transparent and the point arrangement can be set in Format Plot.
Project file format v3
.pivograph files are now saved as schema version 3, recording the analyses, minor ticks, censor ticks, spike styles, and other settings added in this release. Existing v1 and v2 files and example templates migrate automatically when opened, keeping the appearance of survival curve axes and error bars, spike baselines, and parts-of-whole graph sizes. The new specification, JSON Schema, and example files are published, while the v1 and v2 documentation remains available at its existing URLs.
Learn section on the website
Next to Examples and Validation, the new Learn section offers 12 tutorials in English and Chinese: bar graphs with error bars, significance stars, box plots, line graphs with error bars, scatter plots with a regression line, Kaplan–Meier survival curves, t-tests, one-way ANOVA, the difference between SD and SEM, IC50 / EC50 dose-response curves, exporting 300 dpi figures for journals, and importing data from Excel. Each tutorial answers the question in its opening paragraph, illustrates the steps with interface screenshots, and, where a matching example exists, opens the sample data in the editor.
Improvements
Default axes and error bars on survival curves
On new survival curves the X axis starts at time 0 and meets the Y axis, with automatic minor ticks when there are at most three major intervals of 5, 50, or similar. The Y axis runs from 0 to 110 (0 to 1.1 for fractions) with major ticks at 0, 50, and 100 and a minor tick every 5%. Curves always start at time 0, and the start of a curve is no longer cut off when the data begin far from 0. Error bars now default to both directions with narrow caps; other graph types are unchanged. When you switch between percents and fractions in Format Plot, a manually set Y range is converted. Survival curves in existing files keep their axes and error bars when opened.
Survival data read only 1 and 0
In survival tables, a Y of 1 marks an event and 0 marks a censored subject. Rows with any other value, or with a time that is empty, zero, or negative, are left out of the curves and the analysis, and the results report how many rows were ignored. Rows with other values in existing files are no longer drawn on the curves.
t-tests on XY tables
The two data sets of an XY table are two curves over the same X values rather than independent samples, so XY tables no longer offer the t-test; curves can be compared with regression or area under the curve, and the t-test remains available on Column tables. t-tests on XY tables in existing files are removed when the file is opened.
Validation page compared with R value by value
The public validation page (in English and Chinese) now compares each method with R: a summary table at the top lists the compared values and the largest difference for each class of methods, each data set expands to show PivoGraph and R results side by side, and the R script and reference values can be downloaded. The new one-way ANOVA, rank tests, effect sizes, area under the curve, XY correlation, and Kaplan–Meier curves with the survival comparison are all included.
One rule for quantiles
Box plots, the quartile lines of violin plots, median and interquartile range, and descriptive statistics all use the (n + 1)p quantile method, and the 95% confidence interval of the median now uses exact binomial order statistics. Box plots in existing files may change slightly.
Data View shows which column a parts-of-whole graph draws
A parts-of-whole graph draws one column at a time. The top of Data View now names the column being drawn and points to Format Graph › Plotted Data Sets to switch; other columns can still be edited as usual.
Parts-of-whole legends
The Label / Percentage / Value checkboxes on the legend page now sit before their labels, so they no longer read as belonging to the neighboring word. The borders of the legend keys match the outer border on the graph (fill only when the outer border is off) instead of the theme's per-series border colors.
Data symbols drawn above the axes
Data symbols that overlap an axis or its ticks, such as a survival point that drops to 0 or a point at the origin, are now drawn above the axis line; bars remain below the axes.
Double-click an axis to open its settings
Double-clicking the X or Y axis on the graph shows the handles at both ends and opens the page for that axis in Format Axes; double-clicking tick labels or category labels also opens the matching axis.
At least three groups in new tables
New XY tables with replicates, Grouped tables, and Nested tables start with at least three groups, so you can compare groups without adding one first. The import command in the File menu is now named “Import Data (CSV)” to match the export commands.
Fixes
“Graphic size (pt)” now applies to pie charts and other parts-of-whole graphs
The diameter of pie and donut charts, the long side of slices, and the 10 × 10 dot plot are drawn at the size you set, and the plot area grows when the graphic does not fit. Previously the fixed plot area limited the graphic, so a larger setting showed no change in the preview. Values outside 40–300 now show the adjusted value in the field. Existing files keep their sizes when opened.
Vector export keeps circles, arcs, and arrowheads
In SVG, EPS, and PDF exports, circles and ellipses, arcs, and arrowheads of drawing objects are no longer lost, and rounded rectangles keep their rounded corners, matching the screen and raster exports.
No zero-length error bars for single values
When an X value or a group has only one value, mean ± SD and mean ± SEM no longer draw an error bar of length 0; this applies to Column, Grouped, and Nested graphs as well.
Connecting line on mean and error graphs
Checking “Connecting lines” now draws the line between the means.
Significance brackets on before-after graphs stay above the data
Bracket anchors now sit at the highest value of each column; brackets in existing files stay where they were.