Documentation

Trendline overview#

A trendline fits a mathematical relationship to the numerical points in one GraphX dataset. It is calculated after calibration and digitising: GraphX fits the dataset’s graph coordinates, not pixels from the source image. A trendline is therefore a way to analyse the data you extracted, rather than another way to trace a curve.

Start by preparing the points you want to analyse. Datasets explains how points are organised, and Digitising overview covers the extraction workflow.

Create a trendline#

Open Trendlines in the Inspector, select the destination dataset in Datasets, then choose + Add. GraphX creates a new trend row, selects it, and opens its editor. A trend requires at least two finite graph-coordinate points in its source dataset.

In the expanded editor, choose the Dataset source, select a model from Fit, set any model-specific options, then inspect Fit results. Changes update the displayed line and results immediately when the configuration is valid; there is no separate calculation step.

Each row can be selected, expanded or collapsed. The compact row shows its model and source dataset, and includes a colour control, an Export checkbox, reorder controls, and an expand control. − Remove removes the selected trendline without deleting its source data.

Source dataset#

Every trendline references exactly one dataset. The source selector changes which dataset supplies the fitted points; it does not copy those points into the trendline.

When you change the source dataset, GraphX resets the Fit range and Display range to sensible defaults for the new source: calibrated X limits for fitting and the finite data X extent for display. The remaining trendline settings, such as model, colour, visibility, coefficient precision, and any applicable fixed intercept setting, are retained. Review the result after changing source, especially if the new data has a different scale or domain.

Removing a dataset also removes trendlines that reference it. GraphX asks for confirmation when the dataset has points or associated trendlines, so a reference cannot be left pointing to a deleted dataset.

Model types#

GraphX provides five fitted forms:

The equation and coefficient labels shown in the editor use these conventions. For a polynomial, GraphX presents coefficients in descending power order as a, b, c, and so on. The detailed pages explain each model’s assumptions and domain restrictions.

Choose a model because its shape is a reasonable description of the relationship and because its assumptions make sense for the data. A numerical fit can describe a set of points closely without being an appropriate model for explanation or prediction. Understanding R² discusses one useful diagnostic, but a larger R² alone does not choose the correct model.

Read the fit results#

For a valid fit, Fit results shows the general form, the fitted equation, named coefficients, R², Points fitted, Fitted data X range, and the fitting method. Exponential and power fits also show the R² in the transformed space used to estimate their coefficients. Polynomial fits may state Centred/scaled form shown for numerical stability; this is an alternative representation of the same fitted curve, not a different fit.

The result only uses points selected by the fit range. The fitted-data X range reports the smallest and largest X values actually used, which can be narrower than the configured fit range when there are no data points near an endpoint.

Display, colour, and export#

Show trendline on graph controls whether the fitted curve is drawn over the graph. This is separate from the trend row’s Export checkbox. A hidden line remains configured and saved; an unchecked Export box excludes that trendline from PDF export. Keep the source dataset’s Export checkbox enabled as well: the report uses enabled datasets as its source material, so an unchecked dataset cannot contribute its trendline result.

A new trend initially takes the source dataset’s colour. Afterwards, its colour is independent, so changing either colour does not change the other. Trendlines also have their own stable identities and order. Drag the six-dot grip or use the up/down controls to reorder them; order controls the Inspector list and the sequence in which visible lines are drawn.

Enabled trendlines are included in a PDF report, with their model, equation, coefficients, and fit information. The project file stores the trendline configuration, source-dataset reference, colour, visibility, export setting, ranges, and display precision. Coefficients and statistics are recalculated from the restored dataset points rather than treated as separately authoritative saved values.

Coefficient precision#

Coefficient precision controls how coefficients and equations are presented. Choose Significant figures from 1–15 or Decimal places from 0–15; a new trend starts at 4 significant figures. This affects the equation in the editor, Copy equation, and the PDF report. It does not round the internally fitted coefficients or alter the curve drawn on the graph.

Coefficient precision is independent of dataset coordinate Precision. Dataset precision formats data values in tables and exports; trendline precision formats the model description.

A practical workflow#

  1. Digitise and review a dataset.
  2. Add a trendline and confirm its source dataset.
  3. Select the simplest model that is scientifically or analytically appropriate.
  4. Set a fit range if some points should not contribute to the fit.
  5. Inspect the equation, fitted-point count, range, and R².
  6. Set the display range and check any extrapolation warning before relying on the drawn extension.
  7. Keep the line visible or include it in the PDF only when it communicates useful analysis.

For the distinction between the data used to calculate a model and the interval on which it is drawn, see Fit and display ranges.

For the common numerical methods behind the available models, see Regression mathematics.