Purpose

The Analytics Wizard turns a plain-language question about your Cluey / SensingClues data (wildlife sightings, human activity, human-wildlife conflict, points of interest and community work) into a clear analytical report — a chart, a plain-language conclusion and the underlying figures.

Scope

It works on the observation data your teams collect. You choose a period, the data sources (groups) and a topic; the wizard selects a suitable statistical method, checks whether there is enough data, and produces a report. It is a decision-support tool, not a replacement for expert judgement.

How it works

1) You state what you want to analyse (guided, from a predefined list, or as a free-text question). 2) The wizard determines the question type, selects the data and picks a statistical method via a transparent, rule-based decision tree. 3) It checks data sufficiency and computes the result. 4) It shows a chart plus a plain-language conclusion and the technical detail, which you can export.

Where AI is used

AI (a large language model) is used in two places only: to interpret your free-text question into a structured analysis request, and to write the optional plain-language 'Explain the graph' text. All statistical calculations, method selection and data checks are done by deterministic, auditable code — not by AI. AI output is marked with a disclaimer and should always be verified against the data.

Available statistical methods

The wizard chooses automatically from the methods below, grouped by the type of question. You can also switch method manually where alternatives exist.

Descriptive — how much / how many

• Counts and summary — Totals, averages and range of the selected data.

Composition — which type dominates

• Proportions with confidence intervals — Each category's share of the total, with a confidence interval.

• Chi-square goodness-of-fit — Tests whether categories are evenly spread or some dominate.

Comparison — do groups differ

• Kruskal-Wallis group comparison — Tests whether groups differ (non-parametric, 2+ groups).

Trend — change over time

• Mann-Kendall trend test — Tests for a rising or falling trend over time (robust).

• Trend rate (Poisson/neg-binomial GLM) — Estimates the rate of change per period (Poisson / negative binomial).

Association — do two things co-vary

• Correlation (Pearson/Spearman) — Measures whether two things move together (Spearman / Pearson).

Intervention — did an action have an effect

• Before/after comparison (Wilcoxon) — Compares before vs after an intervention date.

• Difference-in-differences — Isolates an intervention's effect against a control group.

• Interrupted time series — Interrupted time series: level and slope change after an intervention.

Spatial — where / co-location

• Hotspot (quadrat) test — Tests whether points cluster into hotspots and locates the busiest area.

• Proximity (nearest-neighbour) — How close one type of point is to another (buffer distance).

• Co-location (Kcross, CSR envelope) — Rigorous test of whether two point patterns co-locate (CSR envelope).