Bellabeat Movement Analysis Tutorial: Steps × METs Integrated Statistical Review

Steps (Volume) × METs (Intensity) — Integrated Statistical Review

This tutorial walks through a structured statistical evaluation of daily steps (movement volume) and MET values (movement intensity).

The goal is simple: Understand behavioral rhythm, stability, and marketing opportunity using descriptive and inferential statistics.

SECTION 1 — Descriptive Analysis: Daily Steps (Volume)

Command Used

indexly analyze-csv avg_daily_steps.csv --show-summary

Dataset Summary — Steps

Interpretation

Steps are:

• Moderately distributed
• Right-skewed (high max vs mean)
• Mostly light-to-moderate daily movers

ASCII Distribution (Conceptual):

Low    ████████████████████
Mid    ███████████████████████████
High   ████

Most users cluster in moderate ranges.

Time-of-Day Pattern (Volume)

From scatter:

00–05  ░░░░░░ (near zero)
07–10  ████
12–18  ███████████
18–19  ███████████████ (peak)
20–23  ███████

Evening peak (~6–7PM).

Weekday Comparison (Steps)

Range ≈ 32 units Relative difference ≈ 20–25%

Highest:

• Tuesday
• Wednesday
• Thursday

Lowest:

• Monday
• Sunday

No extreme weekend spike.

Interim Insight

Steps suggest:

• Stable weekly behavior
• Mild midweek lift
• Strong evening dominance

Already aligning with prior MET analysis.

SECTION 2 — Inferential Statistics (Steps Only)

Weekday Structural Test

indexly infer-csv avg_daily_steps.csv \
  --y avg_daily_steps \
  --group day_of_week \
  --test kruskal

Result:

• H = 7.839
• p = 0.2501

Interpretation:

No statistically significant weekday difference.

Kruskal-Wallis is appropriate when normality assumptions fail [3].

Conclusion: Observed midweek lift is descriptive, not statistically strong.

Weaknesses — Steps Only

• Right-skewed distribution
• High variance
• Aggregation may mask user-level variability
• No seasonal control

Therefore:

This is observational, not causal.

SECTION 3 — Integrated Analysis: Steps × METs

Now we combine volume and intensity.

Merge performed on day_of_week.

Merged rows: 7 (Each weekday becomes one aggregated observation.)

3.1 Pearson Correlation (Primary Test)

indexly infer-csv mets.csv avg_daily_steps.csv \
  --merge-on day_of_week \
  --x mets_pro_mins \
  --y avg_daily_steps \
  --test correlation --use-raw

Result:

Interpretation:

Very strong positive association [1].

As intensity increases, step volume increases.

Statistically significant at α=0.05.

3.2 Spearman Robust Check

indexly infer-csv mets.csv avg_daily_steps.csv \
  --merge-on day_of_week \
  --x mets_pro_mins \
  --y avg_daily_steps \
  --test corr-spearman

Result:

Spearman confirms monotonic relationship [2].

Even under non-normal assumptions, association remains strong.

3.3 OLS Attempt

OLS failed due to:

• n = 7
• Zero variance conditions during bootstrap
• Too few observations for stable regression

This is a sample size limitation, not conceptual failure.

3.4 Structural Weekday Effect (Post-Merge)

indexly infer-csv mets.csv avg_daily_steps.csv \
  --merge-on day_of_week \
  --y avg_daily_steps \
  --group day_of_week \
  --test kruskal

Result:

• H = 6.000
• p = 0.4232

No significant weekday structure after merge.

Cross-Metric Alignment Summary

This is the strongest insight in the dataset.

Core Behavioral Narrative

Users are:

• Routine-driven
• Light-to-moderate movers
• Evening dominant
• Stable across weekdays
• Not weekend warriors

Time-of-day dominates behavior.

Strategic Recommendations for Bellabeat Spring

1️⃣ Position Around Routine

Evidence supports consistency over performance extremes.

Message: “Build sustainable daily habits.”

2️⃣ Evening Engagement Strategy

Peak window: 5PM–8PM.

Actions:

• Push notifications
• Micro-movement prompts
• End-of-day summaries
• Recovery + sleep integration

Evening is behavioral leverage point.

3️⃣ Monday Reset Campaign

Lowest steps occur Monday.

Opportunity: “Monday Restart”

Low risk, high fit.

4️⃣ Full Wellness Loop

Night inactivity supports:

Move → Recover → Sleep → Repeat

Next logical analysis phase: Sleep Data.

Limitations & Responsibility Notice

This analysis:

• Uses aggregated weekday merge (n=7)
• Does not control for individual variation
• Does not include seasonal, demographic, or contextual controls
• Cannot establish causation

Statistical findings are observational.

We strongly recommend reviewing referenced statistical methodology before operational use.

We cannot be held responsible for misinterpretation or misuse of results.

Statistical References

[1] Pearson Correlation — https://statistics.laerd.com/statistical-guides/pearson-correlation-coefficient-statistical-guide.php

[2] Spearman Rank Correlation — https://statistics.laerd.com/statistical-guides/spearmans-rank-order-correlation-statistical-guide.php

[3] Kruskal-Wallis Test — https://statistics.laerd.com/statistical-guides/kruskal-wallis-h-test-statistical-guide.php

Final Integrated Position

Both volume and intensity tell the same story.

Behavior is:

Stable. Habitual. Evening-centered.

Therefore:

Bellabeat Spring should be marketed as a sustainable daily wellness companion for modern women — emphasizing evening engagement, consistency, and balanced recovery.

Next step: Sleep Pattern Analysis — completing the behavioral cycle.

If you’d like to explore the statistical methods in more depth, feel free to check the references section for further reading and background.