paulpas Experimental Design
Comprehensive guide to experimental design in machine learning and data science workflows.
When to Use This Skill
- Solving real-world experimentation & a/b testing problems
- Building machine learning pipelines with experimental design
- Implementing best practices for experimental design
- Optimizing model performance using experimental design techniques
- Learning industry-standard approaches to experimental design
When NOT to Use This Skill
- When using pre-built libraries without understanding underlying concepts
- For toy problems that don't require experimental design rigor
- When domain expertise in specific problem requires different approach
- If your problem doesn't require the complexity this skill provides
Purpose and Key Concepts
Experimental Design is a critical component of the machine learning workflow. This skill covers:
- Theoretical foundations — Mathematical principles and statistical concepts
- Practical implementation — Working code examples and patterns
- Common pitfalls — Mistakes to avoid and how to recover from them
- Best practices — Industry-standard approaches and optimization techniques
Core Workflow
- Understand the problem — Clearly define what you're solving for
- Select approach — Choose the right technique for your data and constraints
- Implement solution — Write clean, tested code following best practices
- Validate results — Verify your implementation with tests and validation
- Optimize performance — Improve efficiency and accuracy incrementally
Implementation Patterns
Pattern 1: Basic Experimental Design
import pandas as pd
import numpy as np
from scipy import stats
def generate_factorial_design(factor_levels: dict) -> pd.DataFrame:
"""Generate a full factorial design matrix from factor levels."""
keys = list(factor_levels.keys())
values = list(factor_levels.values())
grid = np.meshgrid(*values, indexing='ij')
design = pd.DataFrame(np.column_stack([g.ravel() for g in grid]), columns=keys)
return design
def apply_randomization(design: pd.DataFrame, seed: int = 42) -> pd.DataFrame:
"""Randomize the order of experimental runs using a reproducible RNG."""
rng = np.random.default_rng(seed)
indices = rng.permutation(len(design))
return design.iloc[indices].reset_index(drop=True)
def create_blocks(design: pd.DataFrame, n_blocks: int) -> pd.DataFrame:
"""Assign experimental runs to blocks to control for nuisance variables."""
block_assignments = np.repeat(np.arange(n_blocks), len(design) // n_blocks)
remainder = len(design) % n_blocks
if remainder > 0:
block_assignments = np.append(block_assignments, np.arange(remainder))
design['block'] = block_assignments
return design
Pattern 2: Production-Ready Experimental Design
import logging
from typing import Any, Dict, List
import pandas as pd
import numpy as np
from scipy import stats
logger = logging.getLogger(__name__)
class ExperimentalDesign:
"""Production implementation of Experimental Design following ISO 3534 standards."""
def __init__(self, seed: int = 42):
self.seed = seed
self.rng = np.random.default_rng(seed)
def execute(self, data: pd.DataFrame, response_col: str, factors: List[str], n_blocks: int = 1) -> Dict[str, Any]:
"""Execute Experimental Design on data with randomization and blocking."""
if response_col not in data.columns:
raise ValueError(f"Response column '{response_col}' not found in data")
if not all(f in data.columns for f in factors):
raise ValueError(f"All factors must be present in data: {factors}")
design = data[factors].copy()
design = apply_randomization(design, self.seed)
design['block'] = np.repeat(np.arange(n_blocks), len(design) // n_blocks)
remainder = len(design) % n_blocks
if remainder > 0:
design.loc[len(design) - remainder:, 'block'] = np.arange(remainder)
design['response'] = data[response_col].values
if n_blocks > 1:
groups = [group['response'].values for _, group in design.groupby('block')]
f_stat, p_val = stats.f_oneway(*groups)
else:
f_stat, p_val = np.nan, np.nan
results = {
'status': 'success'
'design_matrix': design
'statistics': {'f_statistic': float(f_stat), 'p_value': float(p_val), 'n_runs': len(design)}
'metadata': {'factors': factors, 'blocks': n_blocks, 'seed': self.seed}
}
return results
Best Practices
- ✅ Always validate your implementation on test data
- ✅ Document your assumptions and methodology
- ✅ Use version control for reproducibility
- ✅ Monitor performance metrics in production
- ✅ Periodically review and update your approach
- ✅ Test with edge cases and outliers
- ✅ Log all significant operations for debugging
Common Pitfalls
| Pitfall | Problem | Solution | |
Constraints
MUST DO
- Validate all data preprocessing steps are fit-only on training data, never on validation or test sets
- Implement reproducible pipelines with fixed random seeds and deterministic operations where possible
- Report model performance with confidence intervals via bootstrapping or cross-validation across multiple runs
- Log all experiments with parameters, metrics, and artifacts using MLflow or equivalent tracking system
MUST NOT DO
- Do not evaluate a model on the same data used for training — always hold out a proper test set
- Avoid overfitting to the validation set by limiting hyperparameter search iterations
- Never use features that can only be computed at inference time (look-ahead bias)
- Do not report single-run accuracy without statistical significance testing or error bars
Live References
Authoritative documentation links for this skill's domain. The model follows markdown links at load time to resolve external references and inline content.