Kyūkei Master Hub
SCS2110 | Module 6: Data Analysis & Reliability
👁️ -- opens | 0%

Module 6: Data Analysis, Reliability & Validation

Master Quantitative & Qualitative data analysis methods, statistical testing (ANOVA, t-tests), reliability vs. validity, and data retention rules.

Part 1: Data Analysis Methods Breakdown (Practice Question 5)
M6-Q01 [PRACTICE Q5] [DEFEND & COMPARE]
Discuss different Data Analysis Methods used in Computer Science research. Contrast Quantitative Statistical Analysis with Qualitative Thematic Analysis. [10 Marks]
NUST Model Answer [10 Marks]:

1. Quantitative Analysis Methods:
  • Descriptive Statistics: Summarizing central tendency (mean, median) and dispersion (standard deviation, variance, percentiles) of system performance logs.
  • Inferential Statistics & Hypothesis Testing:
    • Student's t-test: Comparing the mean execution time of two groups (e.g., openEuler vs. Ubuntu baseline).
    • ANOVA (Analysis of Variance): Comparing performance across 3 or more independent system configurations (e.g. comparing latency across 4 cache replacement algorithms).

2. Qualitative Analysis Methods:
  • Thematic Analysis: Systematically coding interview transcripts into recurring themes regarding user experiences or ethical concerns.
  • Grounded Theory Analysis: Iteratively coding data to build an explanatory framework for developer behavior without predefined hypotheses.
M6-Q02 [CLASSIFY & GROUND]
Distinguish between Reliability and Validity in CS research. What is Internal Consistency? [6 Marks]
NUST Model Answer [6 Marks]:
  • Reliability (Consistency): The extent to which an experiment or test yields reproducible, consistent results under identical conditions over time.
  • Validity (Accuracy): The extent to which a test measures what it claims to measure (e.g. ensuring a latency metric accurately reflects network delays rather than client CPU bottlenecks).
  • Internal Consistency: A measure of reliability assessing whether items on a multi-question survey or test suite evaluate the same construct consistently (evaluated statistically using Cronbach's Alpha, where $\alpha \ge 0.70$ is acceptable).
M6-Q03 [CRITIQUE & AUDIT]
DATA AUDIT & RETENTION: According to responsible conduct of research guidelines, what is the mandatory raw data retention period for audit purposes, and why is suppressing conflicting benchmark data considered unethical? [4 Marks]
NUST Model Answer [4 Marks]:
  • Retention Period: Raw data and test logs must typically be retained for 5 years after publication to allow independent verification and institutional auditing.
  • Ethics of Conflicting Data: Suppressing data runs or reporting selective benchmarks that conflict with a proposed hypothesis constitutes a Questionable Research Practice (QRP) and falsification of evidence. All empirical observations must be reported transparently.