Skip to content
Jisan Nuren Siam
Jisan Nuren Siam

Data · Research · Products

All work

Organization & researchSPSS · Statistical review

Confidential Statistical Anomaly Analysis

A multi-stage quantitative review of grouped observational data, designed to find patterns that warrant a closer look. Presented in redacted form.

Overview of a redacted three-stage anomaly analysis: group consistency review, cross-period comparison and temporal pattern analysis, each with its methods and outcome, plus the statistical methods used.
Fig. 01 — Analysis overview
Organization context
FBI · Contributed data & statistical analysis to project work
Role
Data and statistical analysis contributed to project work.
Type
Professional work
Status
Completed analysis · redacted
Tools
SPSS · t-tests · ANOVA · Games-Howell · Pearson correlation
Design
3 linked stages
Data
Grouped observations
Software
SPSS
Output
Review flags

Case study

01Context

Redacted by design

Variables, labels and contextual identifiers are removed for public presentation. Groups, periods and categories are shown generically; no source material is published.

02Stage 1

Group comparison

Descriptive statistics, Welch t-tests, a one-sample test against a benchmark and correlation analysis compared the groups. Large between-group differences were flagged for deeper review.

Stage 1 graphic with anonymized groups: a group means bar chart, pairwise Welch t-test results, a benchmark comparison and a correlation scatter plot.
Fig. 02 — Stage 1 · Group comparison

View full size: Stage 1 · Group comparison (opens in a new tab)

03Stage 2

Cross-period comparison

ANOVA with Levene's test and Games-Howell post-hoc comparisons tested whether the final period differed from earlier ones. It differed significantly from each.

Stage 2 graphic comparing generic periods: a mean difference bar chart, ANOVA and Levene's test results, and Games-Howell comparisons of each period with the final period.
Fig. 03 — Stage 2 · Cross-period comparison

View full size: Stage 2 · Cross-period comparison (opens in a new tab)

04Stage 3

Categories versus time

Differences across categories were not significant; differences across time points were, with the largest shift after the first time point.

Stage 3 graphic: a non-significant comparison across thirteen generic categories next to a significant change across five time points, with post-hoc results.
Fig. 04 — Stage 3 · Temporal patterns

View full size: Stage 3 · Temporal patterns (opens in a new tab)

05Review flags

What the analysis can and cannot say

  • It flags unusual patterns and isolates the periods that need review.
  • It does not determine cause or identify intent.
  • It does not replace a full investigation.
Executive summary of the four review flags, alongside what the analysis supports and what it does not establish: cause, intent or a replacement for a full investigation.
Fig. 05 — Findings and review flags

View full size: Findings and review flags (opens in a new tab)