MODULE I: INTRODUCTION TO STATISTICAL SOFTWARE & DATA MANAGEMENT

Module Architectural Focus: This module delivers an exhaustive, industry-aligned exploration of statistical computing and modern data management protocols for commerce, finance, and empirical business research. It investigates the computational architecture of leading statistical tools—including Microsoft Excel (Data Analysis ToolPak & Power Query), IBM SPSS Statistics (Data View, Variable View & Syntax Architecture), and R/RStudio. The text details Stevens' Four Scales of Measurement, codebook construction, survey response coding, multi-source data ingestion (CSV, Excel, SQL, Web API), data screening, missing value imputation strategies (MCAR, MAR, MNAR), univariate and multivariate outlier detection (Tukey's IQR & Mahalanobis Distance), variable transformations, conditional filtering, and dataset restructuring (Wide vs. Long formats).

1. THE STATISTICAL COMPUTING REVOLUTION IN COMMERCE

The exponential expansion of digital business platforms, enterprise resource planning (ERP) architectures, point-of-sale (POS) terminals, and algorithmic capital markets has generated unprecedented volumes of high-velocity commercial data. In this contemporary ecosystem, reliance on manual, hand-calculated statistical techniques is practically obsolete. Modern commercial survival requires statistical computing environments capable of automating analytical workflows, eliminating arithmetic error, enabling reproducible research, and transforming petabytes of unstructured transactional records into actionable managerial intelligence.

Limitations of Manual Computation

Manual calculations are fundamentally constrained by human fatigue, cognitive error in multi-step equations, inability to handle high-dimensional sample matrices ((N > 100)), and complete lack of an automated audit trail for regulatory compliance.

The Role of Statistical Software

Specialized statistical software provides algorithmic rigor, instantaneous matrix inversion, high-throughput batch processing, dynamic visual dashboards, and automated generation of formal regression and hypothesis testing tables.

Comparative Evaluation of Premier Statistical Environments

Corporate analysts and academic researchers utilize a complementary suite of software applications tailored to specific analytical complexities:

Software Environment	Architectural Paradigm & Interface	Key Strengths & Modules	Commercial & Academic Use Cases
1. Microsoft Excel	Grid-based spreadsheet interface; graphical user interface (GUI) with formula bar and VBA scripting engine.	Ubiquitous enterprise adoption, Power Query ETL, PivotTables, What-If Analysis, Data Analysis ToolPak, Solver add-in.	Financial modeling, operational budgeting, basic descriptive summaries, rapid exploratory data inspection, business dashboarding.
2. IBM SPSS Statistics	Menu-driven GUI paired with dedicated syntax programming language; dual-pane spreadsheet (Data View and Variable View).	Point-and-click statistical menus, automated APA/academic output formatting, robust categorical and survey analysis modules.	Academic dissertations, consumer market research, behavioral human resource analytics, clinical trials, social science surveys.
3. R & RStudio	Open-source, interpreted command-line programming language; integrated development environment (IDE) with console and plot viewer.	Massive CRAN package ecosystem (over 20,000 packages), tidyverse data manipulation, ggplot2 grammar of graphics, R Markdown.	Econometric modeling, quantitative risk management, algorithmic trading analytics, bio-statistics, reproducible academic research.
4. Python (PyData)	General-purpose object-oriented programming language; Jupyter Notebooks and VS Code environments.	Pandas for dataframe manipulation, NumPy for vector algebra, SciPy, Statsmodels, Scikit-learn for machine learning pipelines.	Big data pipelines, automated credit scoring models, predictive artificial intelligence, natural language sentiment processing.

+-----------------------------------------------------------------------------------+ | END-TO-END STATISTICAL DATA PIPELINE ARCHITECTURE | +-----------------------------------------------------------------------------------+ | | | [PHASE 1: DATA ACQUISITION & INGESTION] | | ├── Flat Files: CSV, TSV, Fixed-Width TXT | | ├── Proprietary Sheets: Excel (.xlsx), SPSS (.sav) | | └── Enterprise Databases: SQL, ERP Feeds, Web APIs | | | | | v | | [PHASE 2: VARIABLE METADATA DEFINITION] | | ├── Scale of Measurement: Nominal, Ordinal, Interval, Ratio | | ├── Variable Types: Numeric, String, Date, Currency | | └── Missing Value Coding: System-Missing vs. User-Defined (99, 999) | | | | | v | | [PHASE 3: DATA HYGIENE & CLEANING] | | ├── Validation Checks: Domain ranges, cross-field logic consistency | | ├── Missing Data Diagnostics: MCAR / MAR / MNAR & Imputation | | └── Outlier Screening: Z-scores, Tukey's IQR Fences, Mahalanobis D² | | | | | v | | [PHASE 4: TRANSFORMATION & FEATURE ENGINEERING] | | ├── Variable Computation: Ratios, logs, financial growth metrics | | ├── Recoding: Binning continuous data into ordinal categories | | └── Dataset Merging & Reshaping: Joins, Wide-to-Long Pivoting | +-----------------------------------------------------------------------------------+

2. DATA TYPOLOGY, MEASUREMENT SCALES & METADATA ARCHITECTURE

The mathematical validity of any statistical test is strictly conditional upon the Scale of Measurement of the variables under examination. Applying a parametric test (such as Pearson's correlation or an independent samples (t)-test) to nominal or ordinal data produces methodologically invalid conclusions.

Stevens' Four Levels of Measurement (1946)

MEASUREMENT THEORY

Scale Level	Mathematical Properties	Permissible Mathematical & Statistical Operations	Commercial Business Examples
1. Nominal	Classification only; equivalence / non-equivalence ((=), ( eq)). No natural ordering or magnitude.	Counting, frequencies, mode, percentage distribution, Chi-Square contingency analysis ((chi^2)). Arithmetic mean is meaningless.	Customer Gender (1=Male, 2=Female); Product Category; Geographic Region; Payment Method (Cash, UPI, Card).
2. Ordinal	Classification plus natural rank order ((>), (<)). Distance between successive ranks is unequal or unknown.	Median, percentiles, quartile deviation, Spearman's Rank Correlation ((R)), Mann-Whitney (U), Kruskal-Wallis.	Likert Scales (1=Strongly Disagree to 5=Strongly Agree); Credit Ratings (AAA, AA, A, BBB); Customer Loyalty Tiers.
3. Interval	Ordered scale with known, strictly equal distances between points; arbitrary zero point (zero does not mean total absence).	Addition, subtraction, arithmetic mean, variance, standard deviation, Pearson's (r). Ratios are invalid (80° is not "twice as hot" as 40°).	Calendar Years; Temperature (Celsius/Fahrenheit); Standardized psychological IQ or management aptitude test scores.
4. Ratio	Ordered scale, equal intervals, plus a true absolute physical zero (zero indicates total absence of the attribute).	All mathematical operations: multiplication, division, geometric mean, harmonic mean, coefficient of variation, regression modeling.	Sales Revenue (₹); Production Output (Units); Inventory Turnover; Advertising Budget; Operating Costs; Employee Age.

Variable Classification in Statistical Software

Statistical programs streamline Stevens' four classical scales into operational software formats:

SPSS Variable Architecture

SPSS simplifies measurement into three operational types:

Nominal: Categorical variables without rank order.
Ordinal: Categorical variables with intrinsic order.
Scale: Combines Interval and Ratio scales into a single quantitative numeric metric suitable for parametric analysis.

R Data Class Architecture

R manages data structures through formal object classes:

factor: Qualitative nominal categories.
ordered factor: Ranked ordinal variables.
numeric / double: Continuous real numbers.
integer: Discrete countable values.
character: Plain textual strings.

3. DATA ENTRY, CODING PROTOCOLS & CODEBOOK DEVELOPMENT

The foundation of reliable empirical analysis is structured data entry governed by a formal Codebook. A codebook is a comprehensive technical document that records the location, name, label, variable type, valid response categories, and coding rules for every single item in an investigation.

Standard Codebook Architectural Structure:
A professional commercial research codebook must tabulate:

Variable Identifier (Name): Unique shorthand alphanumeric variable label (e.g., EMP_ID, SATISF_1, ANN_REV).
Variable Label: Comprehensive natural language description representing the exact questionnaire query.
Measurement Scale: Nominal, Ordinal, or Scale/Ratio.
Data Storage Type: Numeric (Width, Decimals), String, Date, or Dollar.
Value Codes & Labels: Numeric mapping for qualitative choices (e.g., $1 = "Retail Banking"$ , $2 = "Wealth Management"$ , $3 = "Corporate Lending"$ ).
Missing Value Designations: Explicit numerical indicators for non-responses (e.g., $99 = "Refused to Answer"$ , $999 = "Not Applicable"$ ).

Handling Multi-Response Questionnaire Items

Commercial surveys frequently include "check-all-that-apply" questions (e.g., "Which banking channels do you utilize regularly?"). In statistical software, these cannot be recorded in a single variable cell. Two industry-standard encoding methodologies exist:

Encoding Methodology	Operational Structure	Software Implementation
1. Multiple Dichotomy Method	Creates a separate binary dummy variable for every single option available in the questionnaire.	Each variable is coded as (1 = Checked (Yes)) and (0 = Unchecked (No)). E.g., `USE_ATM`, `USE_MOBILE`, `USE_BRANCH`. Preferred for SPSS Multi-Response Sets.
2. Multiple Category Method	Creates as many categorical variables as the maximum number of selections permitted to any single respondent.	Variables represent selection slots: `CHANNEL_1`, `CHANNEL_2`, each recording the numeric code of the chosen option.

4. DATA INGESTION, IMPORTING & PLATFORM INTEROPERABILITY

Commercial data rarely originates natively inside specialized statistical software. Analysts must ingest raw files generated by enterprise transactional engines, cloud databases, and survey platforms.

1. Comma-Separated Values (.csv)

The universal open-standard ASCII text format where tabular rows represent observations and columns are delimited by commas. Unaffected by proprietary software versions; universally readable across Excel, SPSS, R, Python, and SQL.

2. Relational Database Feeds (ODBC/SQL)

Statistical environments connect directly to enterprise data warehouses (Oracle, PostgreSQL, Microsoft SQL Server) via Open Database Connectivity (ODBC) drivers, executing SQL queries to ingest filtered records dynamically.

Software Ingestion Workflows

SPSS Data Ingestion Protocol:
• GUI Navigation: File > Import Data > Excel... or Text Data...
• Text Import Wizard Options: Specify delimiter (Comma, Tab, Semicolon), indicate whether variable names are included on the first line, specify decimal separators (period vs. comma), and review data formats.
• SPSS Syntax Command:

GET DATA /TYPE=XLSX /FILE='D:DataCorporate_Branch_Metrics.xlsx' /SHEET=name 'Branch_Data' /CELLRANGE=FULL /READNAMES=ON. EXECUTE.

5. DATA SCREENING, MISSING DATA & OUTLIER DETECTION

Raw empirical datasets invariably contain data entry errors, omissions, uncalibrated sensor records, and extreme anomalies. Statistical modeling without rigorous Data Hygiene leads to biased parameters, inflated standard errors, and corrupted business decisions.

Theoretical Framework of Missing Data (Little & Rubin Taxonomy)

Missing data points must be diagnosed according to their underlying missingness mechanism:

Missingness Mechanism	Statistical Characterization	Methodological Impact & Treatment
1. Missing Completely at Random (MCAR)	The probability of an observation being missing is completely unrelated to any observed or unobserved variable: $P(M mid Y_{ ext{obs}}, Y_{ ext{mis}}) = P(M)$	Benign; reduces statistical power but introduces zero bias. Listwise deletion is statistically valid, or simple mean/median imputation may be applied. Verified via Little's MCAR Test ((p > 0.05)).
2. Missing at Random (MAR)	Missingness is related to observed variables in the dataset, but unrelated to the missing value itself: [ $P(M mid Y_{ ext{obs}}, Y_{ ext{mis}}) = P(M mid Y_{ ext{obs}})$	Listwise deletion yields biased estimates. Requires sophisticated imputation models such as Multiple Imputation by Chained Equations (MICE) or Full Information Maximum Likelihood (FIML).
3. Missing Not at Random (MNAR)	Missingness depends directly on the unobserved value itself (non-ignorable non-response).	E.g., high-income earners refusing to disclose salary. Extremely severe bias; requires specialized pattern-mixture or selection modeling.

Remedial Treatments for Missing Data

Deletion Techniques

• Listwise (Casewise) Deletion: Entire row/case is purged if any variable contains a missing value. Drastically shrinks effective sample size ((N)).
• Pairwise Deletion: Cases are excluded only from specific bivariate calculations involving the missing variable. Produces non-positive definite correlation matrices.

Imputation Techniques

• Mean / Median Substitution: Replaces missing values with variable central tendency. Artificially deflates sample variance and attenuates correlations.
• Regression / KNN Imputation: Predicts missing entries based on observed relationships with other variables.
• Multiple Imputation: Generates (m) plausible datasets, pooling results to capture imputation uncertainty.

Outlier Screening: Univariate & Multivariate Techniques

An outlier is an observation that deviates so markedly from other observations as to arouse suspicion that it was generated by a different mechanism.

Technique 1: Standardized Scores ((Z)-Score Rule)
For approximately normal distributions, standardize data points: [ Z_i = frac{X_i - bar{X}}{s}

Any observation with (|Z_i| > 3.0) (or (|Z_i| > 3.29) for large samples, (p < 0.001)) is flagged as an extreme univariate outlier.

Technique 2: Tukey's Inter-Quartile Range (IQR) Fences
Non-parametric, robust to extreme values. Where (IQR = Q_3 - Q_1):

Inner Fences (Mild Outliers): Outer Fences (Extreme Outliers): [Q_{1} - 1.5 \times I QR, Q_{3} + 1.5 \times I QR] [Q_{1} - 3.0 \times I QR, Q_{3} + 3.0 \times I QR]

]
Technique 3: Mahalanobis Distance ((D^2)) for Multivariate Outliers
Measures multidimensional distance from the centroid of multiple predictor variables: [ D^2 = (X - bar{X})^T Sigma^{-1} (X - bar{X}) ] Evaluated as a Chi-Square variate ((chi^2)) with degrees of freedom equal to the number of variables at (p < 0.001).

6. DATA TRANSFORMATION & DATASET RESTRUCTURING

Raw variables frequently fail to meet parametric distribution assumptions or require mathematical restructuring to align with econometric models.

Mathematical Normalizing Transformations

• Logarithmic ((log_{10} X) or (ln X)): Compresses severe positive right-skewness (e.g., firm revenue, executive compensation).
• Square Root ((√(X))): Stabilizes variance for count data (Poisson-distributed arrival rates).
• Reciprocal ((1/X)): Corrects extreme positive skewness, reversing order of values.

Variable Recoding & Binning

• Recode into Different Variables: Preserves original raw metric while creating a categorized secondary variable (e.g., converting Age in years into 4 distinct demographic brackets).
• Dummy Variable Encoding: Converts an (m)-category nominal variable into (m - 1) binary ((0/1)) indicator variables for regression modeling.

Dataset Merging & Restructuring: Wide vs. Long Data Architecture

Structural Dimension	Wide Data Format	Long (Tidy) Data Format
Record Layout	Each subject/entity occupies exactly one single row. Repeated measurements over time are arranged across multiple columns.	Each measurement occupies a separate row. A subject has multiple rows corresponding to distinct time periods.
Column Structure	`Entity_ID, Revenue_2022, Revenue_2023, Revenue_2024`	`Entity_ID, Year, Revenue`
Software Utilization	Standard cross-sectional analysis, paired (t)-tests, basic multiple linear regression.	Repeated measures ANOVA, linear mixed-effects models, time-series panel data, R ggplot2 visualization.
Conversion Protocol	SPSS: `Data > Restructure > Restructure selected cases into variables`	SPSS: `Data > Restructure > Restructure selected variables into cases` R: `tidyr::pivot_longer()`

7. STEP-BY-STEP PRACTICAL INDUSTRY CASE ILLUSTRATIONS

ILLUSTRATION 1: Setting Up a Comprehensive Survey Codebook & Variable Dictionary
A retail banking conglomerate conducted an empirical Customer Service Experience and Mobile Banking Adoption Survey across 10 commercial branches. The survey administered 6 distinct questions to customers.

Survey Instrument Questions:
1. Customer Account Identification Code.
2. Primary Banking Branch Type (Choices: Metro Urban, Semi-Urban, Rural).
3. Customer Age in completed years.
4. Customer Gender (Choices: Male, Female, Prefer not to say).
5. Mobile Banking Application Satisfaction: "The mobile app is user-friendly and responsive" (Choices: 1=Strongly Disagree, 2=Disagree, 3=Neutral, 4=Agree, 5=Strongly Agree).
6. Annual Average Deposit Balance (in thousands of text{text{Rs. } }).

Required: Construct a formal SPSS Variable Dictionary and Data Codebook specifying Variable Name, Type, Width, Decimals, Label, Values Coding Scheme, Missing Value Codes, and Measurement Scale.

Solution & Step-by-Step Codebook Architecture:

Variable Name	Type	Width/Dec	Variable Label	Value Labels & Coding Scheme	Missing	Measure
`CUST_ID`	Numeric	8, 0	Customer Unique Account ID	None (Continuous Unique Key)	None	Nominal
`BRANCH_TYP`	Numeric	2, 0	Primary Banking Branch Category	1 = Metro Urban 2 = Semi-Urban 3 = Rural	99	Nominal
`CUST_AGE`	Numeric	3, 0	Customer Age in Completed Years	None (Ratio continuous age)	999	Scale
`GENDER`	Numeric	1, 0	Self-Identified Customer Gender	1 = Male 2 = Female 3 = Prefer not to say	9	Nominal
`APP_SATISF`	Numeric	1, 0	Mobile App Usability & Satisfaction	1 = Strongly Disagree 2 = Disagree 3 = Neutral 4 = Agree 5 = Strongly Agree	9	Ordinal
`ANN_DEP`	Numeric	8, 2	Annual Average Deposit Balance (text{text{Rs. } } '000)	None (Direct monetary scale metric)	-99.00	Scale

SPSS Syntax Implementation for Automated Dictionary Generation:

VARIABLE LABELS CUST_ID 'Customer Unique Account ID' BRANCH_TYP 'Primary Banking Branch Category' CUST_AGE 'Customer Age in Completed Years' GENDER 'Self-Identified Customer Gender' APP_SATISF 'Mobile App Usability & Satisfaction' ANN_DEP 'Annual Average Deposit Balance (INR Thousands)'. VALUE LABELS BRANCH_TYP 1 'Metro Urban' 2 'Semi-Urban' 3 'Rural' 99 'Not Disclosed' /GENDER 1 'Male' 2 'Female' 3 'Prefer not to say' 9 'Missing' /APP_SATISF 1 'Strongly Disagree' 2 'Disagree' 3 'Neutral' 4 'Agree' 5 'Strongly Agree' 9 'Unanswered'. MISSING VALUES BRANCH_TYP (99) / CUST_AGE (999) / GENDER (9) / APP_SATISF (9) / ANN_DEP (-99.00). VARIABLE LEVEL CUST_ID, BRANCH_TYP, GENDER (NOMINAL) /APP_SATISF (ORDINAL) /CUST_AGE, ANN_DEP (SCALE). EXECUTE.

ILLUSTRATION 2: Outlier Screening Using Tukey's IQR Fences & Winsorization
A regional credit cooperative inspected the daily withdrawal transaction amounts (in thousands of text{text{Rs. } }) across a sample of 12 corporate accounts on a festive operating day:
Raw Sample Values: 15, 18, 20, 22, 25, 26, 28, 30, 32, 35, 72, 110

Required:
1. Compute the First Quartile ((Q_1)), Median ((M)), Third Quartile ((Q_3)), and the Inter-Quartile Range ((IQR)).
2. Establish Tukey's Lower and Upper Inner Fences and Outer Fences.
3. Identify any mild or extreme outliers in the dataset.
4. Perform a 90% Winsorization to remediate extreme values without losing sample points.

Solution & Step-by-Step Working Notes:

Sorted dataset ((N = 12)): 15, 18, 20, 22, 25, 26, 28, 30, 32, 35, 72, 110.

Step 1: Compute Quartiles:
• Position of First Quartile (Q_1 = (N + 1) / (4) = (13) / (4) = 3.25^{text{th}}) value.
(Q_1 = 3^{text{rd}} text{ value} + 0.25 × (4^{text{th}} - 3^{text{rd}}) = 20 + 0.25(22 - 20) = 20 + 0.50 = mathbf{20.50}).

• Position of Median (M = (N + 1) / (2) = 6.5^{text{th}}) value: ((26 + 28) / (2) = mathbf{27.0}).

• Position of Third Quartile (Q_3 = (3(N + 1)) / (4) = (39) / (4) = 9.75^{text{th}}) value.
(Q_3 = 9^{text{th}} text{ value} + 0.75 × (10^{text{th}} - 9^{text{th}}) = 32 + 0.75(35 - 32) = 32 + 2.25 = mathbf{34.25}).

• Inter-Quartile Range ((IQR)): [ IQR = Q_3 - Q_1 = 34.25 - 20.50 = mathbf{13.75}

Step 2: Establish Tukey's Fencing Boundaries:
• Inner Fences:

Lower Inner Fence = Q_{1} - 1.5 (I QR) = 20.50 - 1.5 (13.75) = 20.50 - 20.625 = - 0.125

Upper Inner Fence = Q_{3} + 1.5 (I QR) = 34.25 + 1.5 (13.75) = 34.25 + 20.625 = 54.875

• Outer Fences:

Upper Outer Fence = Q_{3} + 3.0 (I QR) = 34.25 + 3.0 (13.75) = 34.25 + 41.25 = 75.50

Step 3: Diagnostic Outlier Identification:
• Observation 72 lies between the Upper Inner Fence (54.875) and Upper Outer Fence (75.50). It is classified as a Mild Outlier.
• Observation 110 lies strictly beyond the Upper Outer Fence (75.50). It is classified as an Extreme Outlier.

Step 4: Remedial Winsorization:
Under Winsorization, rather than deleting outliers (which causes loss of statistical degrees of freedom), the extreme values are replaced by the nearest acceptable boundary percentile values (here, the upper boundary value before outlier threshold, (X = 35)):
• Winsorized Series: 15, 18, 20, 22, 25, 26, 28, 30, 32, 35, 35, 35.
Managerial Result: Original mean was $X ˉ = \frac{433}{12} = 36.08$ . The Winsorized mean is $X ˉ_{w} = \frac{316}{12} = 26.33$ , demonstrating how Winsorization successfully eliminates distortion while retaining the full sample size ((N = 12)).

ILLUSTRATION 3: Data Transformation, Conditional Filtering & Merging
A multi-branch retail organization maintains two independent monthly departmental data tables in CSV format:
• Table A (Operations): Branch_ID, Footfall, Gross_Sales (in lakhs ₹), Cost_of_Goods (in lakhs ₹).
• Table B (Quality): Branch_ID, Customer_Rating (1 to 10 scale), Audit_Score (out of 100).

Required:
1. Write the statistical workflow and pseudo-code/syntax to perform an Inner Join merging Table A and Table B on the primary key Branch_ID.
2. Formulate a transformation computing Gross_Margin_Pct defined as:

Gross_Margin_Pct = (\frac{Gross_Sales - Cost_of_Goods}{Gross_Sales}) \times 100

3. Formulate the conditional filter selecting only branches with Gross_Margin_Pct

\geq 30%

AND Customer_Rating

\geq 8.0

for an executive performance reward.

Solution & Procedural Workflow:

1. Database Merge (Join) Protocol:
Both files must be sorted in ascending order by the key matching variable Branch_ID prior to merging.

* SPSS Syntax for Merging Datasets. MATCH FILES /FILE='D:DataTable_A_Operations.sav' /TABLE='D:DataTable_B_Quality.sav' /BY Branch_ID. EXECUTE. * R / tidyverse Syntax Equivalent. library(dplyr) merged_data <- inner_join(table_a, table_b, by = "Branch_ID")

2. Variable Transformation (Feature Engineering):

* SPSS Syntax for Computing Margin. COMPUTE Gross_Margin_Pct = ((Gross_Sales - Cost_of_Goods) / Gross_Sales) * 100. VARIABLE LABELS Gross_Margin_Pct 'Branch Gross Profit Margin Percentage'. EXECUTE. * R Equivalent. merged_data <- merged_data %>% mutate(Gross_Margin_Pct = ((Gross_Sales - Cost_of_Goods) / Gross_Sales) * 100)

3. Conditional Selection and Filtering:

* SPSS Syntax for Executive Incentive Cohort. USE ALL. COMPUTE filter_$ = (Gross_Margin_Pct >= 30.0 & Customer_Rating >= 8.0). VARIABLE LABELS filter_$ 'Gross_Margin_Pct >= 30 & Rating >= 8.0 (FILTER)'. VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected for Executive Reward'. FORMATS filter_$ (f1.0). FILTER BY filter_$. EXECUTE. * R Equivalent. executive_reward_cohort <- merged_data %>% filter(Gross_Margin_Pct >= 30.0, Customer_Rating >= 8.0)

8. COMPREHENSIVE SELF-ASSESSMENT MODULE

SECTION A: CONCEPTUAL & TECHNICAL MCQS

Select the most appropriate option. Complete solution key with technical rationale follows.

1. Which scale of measurement possesses an arbitrary zero point where ratios between numbers are mathematically invalid?
(a) Nominal (b) Ordinal (c) Interval (d) Ratio

2. In SPSS Statistics, the two distinct operational views accessible in the Data Editor window are:
(a) Code View and Layout View (b) Data View and Variable View
(c) Table View and Chart View (d) Input View and Output View

4. In Tukey's boxplot fencing method, an observation is formally classified as an extreme outlier if it falls beyond:
(a) $Q_{1} - 1.5 \times I QR$ (b) $Q_{3} + 1.5 \times I QR$
(c) $Q_{3} + 3.0 \times I QR$ (d) $X ˉ + 2.0 \times σ$

5. Which technique for handling outliers replaces extreme values with the nearest non-outlier percentile boundary value without discarding records?
(a) Listwise Deletion (b) Winsorization (c) Mean Imputation (d) Casewise Trimming

6. In a dataset of customer survey responses, Likert-scale items (1=Strongly Disagree to 5=Strongly Agree) are fundamentally measured on which scale?
(a) Nominal (b) Ordinal (c) Interval (d) Ratio

7. What is the primary statistical consequence of substituting missing values with the arithmetic mean of the variable?
(a) Inflates the standard deviation (b) Artificially reduces variance and attenuates correlations
(c) Converts the data to a normal curve (d) Eliminates all multicollinearity

8. To restructure a dataset from 'Wide' format (multiple quarterly columns) to 'Long' format (single column with time index) in R, which command is utilized?
(a) pivot_wider() (b) pivot_longer() (c) merge() (d) bind_rows()

9. When importing delimited text files into statistical software, which character is the universal delimiter in standard .csv files?
(a) Semicolon (b) Tab (c) Comma (d) Pipe symbol

Answers & Conceptual Rationale for Section A:
1. (c) Interval: Interval scales have equal distances between units but an arbitrary zero (e.g., Celsius temperature), making ratios invalid.
2. (b) Data View and Variable View: The core dual architecture of the SPSS Statistics data editor.
3. (c) MCAR: Missing Completely at Random implies missingness is an entirely random stochastic process.
4. (c) $Q_{3} + 3.0 \times I QR$ : Points beyond 3 times the IQR outside the quartiles are defined by Tukey as extreme outliers.
5. (b) Winsorization: Capping extreme observations at specific percentile fences is known as Winsorization.
6. (b) Ordinal: Likert items have clear directional ordering, but distances between points cannot be mathematically guaranteed to be equal.
7. (b): Mean imputation adds identical mean values, artificially concentrating observations at the center and deflating true dispersion.
8. (b) pivot_longer(): The standard tidyverse verb to lengthen data by increasing rows and decreasing columns.
9. (c) Comma: CSV stands explicitly for Comma-Separated Values.
10. (b) Mahalanobis Distance ((D^2)): Evaluates multidimensional distance of an observation from the joint centroid of multiple variables.

SECTION B: ANALYTICAL & THEORETICAL ESSAY QUESTIONS

Critically evaluate Stevens' Four Levels of Measurement. Why is identifying the measurement scale a mandatory prerequisite before selecting a statistical hypothesis test? Illustrate with commercial examples.
Differentiate between Listwise Deletion, Pairwise Deletion, and Multiple Imputation. Under what missing data mechanisms (MCAR, MAR, MNAR) is each technique methodologically appropriate?
Describe the architecture of the SPSS Variable View. Explain the functional purpose of the Values, Missing, and Measure columns in structuring a corporate survey database.
Explain the mathematical rationale behind Tukey's IQR Fences for outlier detection. Why is the IQR fence preferred over standard (Z)-scores when dealing with skewed financial distributions?
Contrast the Wide and Long data formats. Outline two practical business analysis scenarios where restructuring data from wide to long is essential.

SECTION C: ADVANCED PRACTICAL BUSINESS SCENARIOS

Scenario 1: Codebook Design for an E-Commerce Consumer Loyalty Study
An e-commerce marketplace conducts a customer shopping satisfaction study across 5,000 verified buyers. The questionnaire collects: (a) Transaction Invoice Number, (b) Device Type (Desktop, Mobile App, Tablet), (c) Annual Household Income in ₹, (d) Customer Satisfaction Rating on a 7-point Likert scale, (e) Primary Product Category Purchased (Electronics, Fashion, Groceries, Home Goods), (f) Net Promoter Score recommendation likelihood (0 to 10), and (g) Whether the user has opted for Prime Membership (Yes/No).

Required:
1. Design a comprehensive Codebook table specifying: Variable Name, Label, Data Type, Measurement Scale, Value Coding Labels, and Missing Value rules.
2. Identify which variables are permissible for arithmetic mean calculations and which must be limited to frequencies and modes.
3. Write the exact SPSS Syntax code commands to instantiate this variable structure in an active dataset.

Scenario 2: Data Hygiene & Outlier Treatment in Loan Appraisal Data
A microfinance institution records the monthly earnings (in thousands of ₹) of 10 micro-entrepreneurs applying for business credit: 12, 14, 15, 16, 18, 20, 22, 25, 45, 95.

Required:
1. Calculate the mean, standard deviation, and median of the raw earnings series.
2. Compute the (Z)-score for each applicant and identify any univariate outliers using the (|Z| > 2.0) criterion.
3. Compute (Q_1, Q_3, IQR) and determine Tukey's Inner and Outer Fences.
4. Compare the diagnostic findings of the (Z)-score method versus the IQR fence method on this positively skewed distribution.
5. Formulate a managerial policy recommending whether the credit officer should trim, winsorize, or retain the extreme earnings figures during creditworthiness evaluation.