Interval Plot

On this page
  1. Purpose
  2. Key Capabilities
  3. Statistical Methods
  4. Input Fields & Options
  5. Output Descriptions
  6. Worked Example

Purpose

The Interval Plot shows the mean of a response for each level of a grouping variable, with an interval bar around each mean: a confidence interval for the mean or a multiple of its standard error. It gives a quick picture of which groups (suppliers, machines, shifts) are likely to differ and how precisely each mean is known, before a formal test such as One-Way ANOVA.

Key Capabilities

  • One mean and interval bar per group, with the means joined by a line, plus a table of N, Mean, StDev and the interval limits.
  • Confidence interval at any level (default 95%), or mean ± a chosen number of standard errors.
  • Bonferroni option, so that all the intervals together hold the chosen confidence level.
  • Two-sided intervals, or upper or lower one-sided bounds.
  • Each group's own standard deviation, or a pooled standard deviation across groups.

Statistical Methods

Rows with a blank response or group are dropped. For each group i the tool computes the size nᵢ, the mean ȳᵢ and the sample standard deviation sᵢ, and the standard error SEᵢ = sᵢ / √nᵢ with nᵢ − 1 degrees of freedom. With Pool error across groups ticked, every group uses the pooled standard deviation sₚ = √[Σ(nᵢ − 1)sᵢ² / (N − k)], with N − k degrees of freedom (N observations, k groups), so all groups of the same size get bars of the same length.

For a confidence interval at level C, α = 1 − C/100 and the bar is ȳᵢ ± t × SEᵢ, with t the 1 − α/2 quantile of Student's t for a two-sided interval and the 1 − α quantile for a one-sided bound. With Bonferroni, α is divided by the number of groups k, which widens every bar so that the chance that all k intervals cover their true means is at least C. For the Standard error type the bar is ȳᵢ ± m × SEᵢ, where m is the Multiple (default 1); such bars are not confidence intervals.

Non-overlapping 95% intervals suggest the means differ; overlapping ones do not prove they are equal. The plot is a guide, and One-Way ANOVA with a comparison method is the formal test.

Input Fields & Options

Field

Description

Y (response)

Required. Numeric response column.

X (categorical grouping)

Required. Column whose levels form the groups.

Type of interval

Confidence interval (default) or Standard error.

Level

Confidence level in percent, default 95. Used for Confidence interval only.

Bonferroni

Off by default. Controls the simultaneous confidence level of all the intervals. Confidence interval only.

Multiple

Number of standard errors, default 1. Used for Standard error only.

Side

Two-sided (default), Upper one-sided (bars above the mean only) or Lower one-sided (bars below the mean only).

Pool error across groups

Off by default. Uses the pooled standard deviation and degrees of freedom instead of each group's own.

Data Options: Subset (optional)

Exclude matching rows (default) or Include only matching rows; Row numbers (e.g. 3,7,12-15) and/or a Condition (e.g. Supplier != "Supplier D") select the rows.

interval-plot-inputs


Figure 1 — Options for the example: Strength_MPa by Supplier, two-sided 95% confidence intervals with each supplier's own standard deviation.

Output Descriptions

  • Interval Plot of <Y> vs <X> — each group mean with its interval bar, the means joined by a line; the subtitle states the interval type and whether individual or pooled standard deviations were used.
  • Means — per group: N, Mean, StDev, Lower and Upper limits of the interval (shown as * for the side not calculated with a one-sided bound).
  • PDF export via Download.

interval-plot-output


Figure 2 — 95% confidence intervals for the mean tensile strength of four suppliers, eight coupons each, with the table of means and limits.

Worked Example

Four suppliers deliver material to the same tensile-strength specification. Eight coupons from each supplier are tested, and the quality engineer wants a first picture of which suppliers differ. Y (response) = Strength_MPa, X (categorical grouping) = Supplier, Type of interval = Confidence interval at Level 95, two-sided, without pooling.

Supplier B has the highest mean, 152.895 MPa (95% interval 151.213 to 154.577), and the smallest spread (StDev 2.01). Supplier D has the lowest, 141 MPa (137.097 to 144.903); its interval does not overlap B's or C's (145.094 to 151.708), so D is probably weaker than both. Supplier A (143.332 to 150.493) overlaps C and D and cannot be separated from them by eye.

Because four intervals are compared at once, ticking Bonferroni gives a fairer picture: the bars widen (Supplier D becomes 135.494 to 146.506, Supplier B 150.523 to 155.267), and D and B still do not overlap. Pooling the error gives every supplier the same half-width of about 2.80 MPa. One-Way ANOVA on the same two columns gives p < 0.001, so at least one supplier differs; the engineer would follow up with Supplier D on its process before the next delivery.