AI Skill Report Card

Conducting Scientific Research

B+75·Jun 14, 2026·Source: Web

Scientific Research Methodology

10 / 15
Python
# Research pipeline example research_pipeline = { "question": "How does X affect Y?", "hypothesis": "X increases Y through mechanism Z", "methods": ["experimental_design", "data_collection", "statistical_analysis"], "analysis": "statistical_tests + visualizations", "conclusion": "evidence_based_findings" }
Recommendation
Make Quick Start more actionable with a specific research problem and complete workflow example rather than just a generic pipeline structure
15 / 15

Progress:

  • Literature Review: Search databases (PubMed, Google Scholar, arXiv)
  • Research Question: Formulate specific, testable hypothesis
  • Methodology Design: Choose appropriate experimental/analytical methods
  • Data Collection: Gather primary/secondary data systematically
  • Data Analysis: Apply statistical methods and create visualizations
  • Results Interpretation: Draw evidence-based conclusions
  • Paper Writing: Structure findings in academic format
  • Peer Review: Validate methodology and conclusions

Literature Review Process:

  1. Define search terms and databases
  2. Screen abstracts for relevance
  3. Full-text review of selected papers
  4. Extract key findings and methodologies
  5. Identify research gaps

Experimental Design:

  1. Define variables (independent, dependent, control)
  2. Determine sample size using power analysis
  3. Select randomization and blinding strategies
  4. Plan data collection protocols
  5. Establish statistical analysis plan
Recommendation
Reduce length by removing verbose explanations (e.g., the detailed Literature Review and Experimental Design sub-sections could be condensed)
15 / 20

Example 1: Hypothesis Formation Input: "Climate change effects on crop yields" Output:

  • Research Question: "How do increasing temperatures affect wheat yield in temperate regions?"
  • Hypothesis: "Rising temperatures above optimal range (15-20°C) will decrease wheat yields by 10-15% per degree increase"
  • Variables: Temperature (independent), wheat yield (dependent), rainfall/soil (controls)

Example 2: Data Analysis Input: Experimental dataset with treatment groups Output:

Python
# Statistical analysis framework import scipy.stats as stats import pandas as pd # Descriptive statistics summary_stats = data.groupby('treatment').agg({ 'outcome': ['mean', 'std', 'count'] }) # Hypothesis testing t_stat, p_value = stats.ttest_ind(control_group, treatment_group) effect_size = (treatment_mean - control_mean) / pooled_std

Example 3: Research Paper Structure Input: Completed analysis results Output:

1. Abstract (250 words): Objective, methods, results, conclusion
2. Introduction: Background, literature gap, research question
3. Methods: Study design, participants, procedures, analysis
4. Results: Descriptive stats, main findings, figures/tables
5. Discussion: Interpretation, limitations, implications
6. References: APA/journal format
Recommendation
Provide more concrete input/output examples - Example 1 shows good specificity but Examples 2-3 are too template-like rather than showing actual research scenarios

Research Design:

  • Use PICOT framework (Population, Intervention, Comparison, Outcome, Time)
  • Pre-register study protocols to prevent bias
  • Calculate required sample sizes before data collection
  • Include positive and negative controls where appropriate

Data Management:

  • Use version control for data and analysis code
  • Document all preprocessing steps
  • Implement data validation checks
  • Store raw data separately from processed data

Statistical Analysis:

  • Check assumptions before applying tests
  • Use appropriate multiple comparison corrections
  • Report effect sizes alongside p-values
  • Include confidence intervals for estimates

Writing:

  • Follow journal-specific formatting guidelines
  • Use active voice and clear, concise language
  • Include all necessary details for reproducibility
  • Cite recent and relevant literature

Study Design Issues:

  • Correlation vs Causation: Don't imply causation from observational data
  • Sample Bias: Ensure representative sampling for target population
  • Confounding Variables: Control for alternative explanations
  • Underpowered Studies: Calculate sample size requirements early

Analysis Errors:

  • P-hacking: Don't fish for significant results by trying multiple tests
  • Cherry-picking: Report all analyses performed, not just significant ones
  • Inappropriate Tests: Match statistical tests to data type and distribution
  • Missing Data: Handle systematically, don't just delete cases

Reporting Problems:

  • Selective Reporting: Include non-significant results
  • Overgeneralization: Stay within bounds of your sample and methods
  • Incomplete Methods: Provide sufficient detail for replication
  • Unsupported Claims: Base conclusions strictly on presented evidence

Reproducibility:

  • Code Documentation: Comment analysis code thoroughly
  • Data Availability: Share data when possible per journal policies
  • Method Transparency: Report all steps, including failed approaches
  • Version Control: Track changes to analysis and manuscripts
0
Grade B+AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
10/15
Workflow
15/15
Examples
15/20
Completeness
15/20
Format
15/15
Conciseness
10/15