ARADMANAMNAOON/

DATA SCIENCE / REAL-WORLD RESEARCH / CASE STUDY

Messy data.
Clearer thinking.

I took a complicated psychology dataset, checked whether the numbers could be trusted, and uncovered a surprising finding that a standard chart would have missed.

See what I discoveredDownload the research
An exploratory audit—not proof of a psychological group effect
240 records86distinct answer patterns
Fewer unique answers than expected One dot for each distinct pattern
FROM ROWS TO REASONS01 / 03

01 / THE PROBLEM

The spreadsheet looked ordinary.
It wasn't.

At first glance, the study had two equally sized groups and hundreds of answers. Instead of immediately testing which group scored higher, I checked whether the underlying responses made sense.

ROWS EXAMINED

240Recorded questionnaire rows

DISTINCT ANSWER PATTERNS

86Many records repeated the same answers

MATCHING CROSS-GROUP PAIRS

119/120Almost every aligned pair answered 76 items identically
✳

What this means: Different rows don't always mean independent answers. Spotting that distinction is part of the analysis—not a footnote.

02 / THE DISCOVERY

A big difference.
Until I checked.

The saved scores seemed to separate the two groups. But when I checked the actual answers behind those totals, the apparent gap almost disappeared.

CLICK TO EXPLOREINTERACTIVE / 01

DIFFERENCE BETWEEN THE TWO CODED GROUPS

4.63

The stored total suggests a visible gap between the groups.

↳

Switch the calculation above to see why checking the original items matters.

GROUP AVERAGEAMBIGUITY-TOLERANCE ITEMS
fire40.45
non35.83

Both values are descriptive averages. The underlying items may require reverse scoring, so the direct sum is a diagnostic—not an official scale score.

!

The insight isn't that one group is better.

It's that the result changes dramatically when we question how it was calculated. A responsible analysis finds that issue before telling people what the study supposedly proves.

03 / WHAT I DELIVERED

Not just numbers.
A better question.

My job was to turn an untidy SPSS file into understandable evidence, including what needed correcting before anyone could trust a stronger claim.

01
⌕

Found the hidden problem

Repeated answers, inconsistent totals, and 31 invalid item responses would have distorted a simple story.

02
↗

Tested the apparent result

I compared stored totals with the original questionnaire answers instead of assuming the exported scores were correct.

03
✦

Made the finding understandable

I translated technical checks into a clear visual explanation, while keeping the limits and complete analysis open for review.

IMPORTANT CONTEXT

This is an exploratory data-integrity case study. The original study design, group meanings, and official scoring rules still need confirmation. No clinical, causal, or independent-group claim is made.

THE EVIDENCE / OPEN FOR REVIEW

Want the full story?

Every statistical table, assumption check, method, and limitation is explained in the English and Persian reports. The R analysis source is provided for technical review.

EN
Complete English reportStatistical methods, results, interpretations, and appendices
PDF ↘Word ↘
فا
Complete Persian reportFull right-to-left report with explanations and tables
PDF ↘Word ↘
R
Analysis source code (R)Complete R analysis script · execution not independently verified
R code ↘

The participant-level SPSS source is named raw.sav. It is withheld from public download pending appropriate data-sharing permission.