Back to Agents

Category E: Analysis Agents

Paradigm-appropriate analytical strategies and implementation

Analysis agents provide comprehensive support for quantitative, qualitative, and mixed methods data analysis. They generate executable code, guide statistical decisions, and ensure robust analytical practices across all research paradigms.

Core Principle

Paradigm-specific analysis strategies with reproducible code generation

Quantitative Analysis & Code Gen

E1

Guide statistical analysis selection, assumption checking, interpretation, code generation (R, Python, SPSS, Stata, Mplus), and sensitivity analysis

HIGH • Opus
Enhanced VS 3-Phase
CP_ANALYSIS_PLAN

Capabilities

  • •Descriptive statistics and exploratory data analysis
  • •Inferential statistics (t-test, ANOVA, regression, SEM)
  • •Assumption checking (normality, homogeneity, independence)
  • •Effect size calculation and interpretation
  • •Multilevel and longitudinal modeling
  • •Meta-analysis techniques
  • •Code generation: R (tidyverse, lavaan, lme4, psych), Python (pandas, statsmodels, scipy, scikit-learn)
  • •Code generation: SPSS syntax, Stata commands, Mplus syntax, NVivo/ATLAS.ti queries
  • •Sensitivity analysis: outlier influence, model specification alternatives, multiverse/specification curve
  • •Robustness checks: assumption relaxation, subgroup analysis, meta-analysis sensitivity

VS Process

Phase 1: Understand research design and data structure | Phase 2: Modal analysis awareness (e.g., t-test/ANOVA dominance) | Phase 3: Present differentiated analytical strategies with executable code and sensitivity checks

Example

Input
"Compare learning outcomes across 3 groups with pre-post measures"
Output
Modal (T≈0.8): Repeated measures ANOVA | Direction A (T≈0.5): ANCOVA with pretest covariate | Direction B (T≈0.3): Growth curve modeling with group × time interaction | Recommended: ANCOVA controls for baseline differences, SEM for mediation paths | Code: R/SPSS/Stata generated | Sensitivity: specification curve across model variants
Triggers:statistical analysisANOVAregressionSEMt-testchi-squaremultilevel modelingR codePython codeSPSS syntaxStata codeMplusanalysis scriptcode generationsensitivity analysisrobustness checkspecification curve

Qualitative Coding Specialist

E2

Systematic coding strategy development for thematic analysis, grounded theory, and other qualitative approaches

HIGH • Opus
Enhanced VS 3-Phase

Capabilities

  • •Codebook development (deductive, inductive, hybrid)
  • •Thematic analysis (Braun & Clarke framework)
  • •Grounded theory coding (open, axial, selective)
  • •Content analysis and discourse analysis
  • •Saturation assessment strategies
  • •CAQDAS software guidance (NVivo, ATLAS.ti, MAXQDA)

VS Process

Phase 1: Identify coding approach | Phase 2: Avoid mechanical coding, encourage interpretive depth | Phase 3: Present coding strategies with theoretical grounding

Example

Input
"Code interview data about teacher AI experiences"
Output
Approach A (T≈0.6): Thematic analysis with 6-phase process | Approach B (T≈0.4): Grounded theory - open coding for emergent concepts | Approach C (T≈0.2): Critical discourse analysis examining power dynamics | Codebook structure: Descriptive → Interpretive → Pattern codes
Triggers:qualitative codingthematic analysiscodingNVivoATLAS.tiMAXQDAgrounded theory coding

Mixed Methods Integration Specialist

E3

Design integration strategies for combining quantitative and qualitative data to generate meta-inferences

HIGH • Opus
Enhanced VS 3-Phase
CP_INTEGRATION_STRATEGY

Capabilities

  • •Joint display creation (comparison, integration, synthesis)
  • •Data transformation (quantitizing, qualitizing)
  • •Convergence/divergence analysis
  • •Meta-inference development
  • •Integration timing (concurrent vs. sequential)
  • •Legitimation strategies for mixed methods

VS Process

Phase 1: Assess integration purpose | Phase 2: Identify modal integration (simple side-by-side) | Phase 3: Present creative integration strategies with methodological rigor

Example

Input
"Integrate survey data (N=300) and interview data (n=20) on AI attitudes"
Output
Strategy A (T≈0.6): Joint display comparing themes to scale scores | Strategy B (T≈0.4): Quantitize themes → cluster analysis → qual explanation of clusters | Strategy C (T≈0.2): Configurational analysis (QCA) blending both strands | Meta-inference: How do lived experiences explain statistical patterns?
Triggers:mixed methods integrationjoint displaymeta-inferencedata transformationconvergencedivergence

Checkpoint Integration

Analysis agents use checkpoints to ensure methodological rigor:

CP_ANALYSIS_PLAN

E1

Statistical analysis plan approved before execution

CP_INTEGRATION_STRATEGY

E3

Mixed methods integration strategy confirmed

Paradigm Coverage

Analysis agents adapt to your research paradigm:

Quantitative

E1

Statistical analysis, code generation, sensitivity/robustness checks

Qualitative

E2

Coding strategies, CAQDAS support, trustworthiness

Mixed Methods

E1E2E3

Full pipeline with integration strategies

Multi-Language Code Generation

E1 generates production-ready code across platforms:

R

tidyverse, lavaan, lme4

Statistical modeling and visualization

Python

pandas, statsmodels, scipy

Data analysis and machine learning

SPSS

Syntax

Point-and-click alternative with reproducibility

Stata

Command syntax

Panel data and econometric models

Mplus

SEM syntax

Structural equation modeling, LCA, multilevel

NVivo

Query language

Qualitative coding and retrieval

Typical Analysis Workflow

1
E1 / E2

Select analysis approach, generate code, and design sensitivity checks (paradigm-specific)

CP_ANALYSIS_PLAN
2
E3

Integration strategy (if mixed methods)

CP_INTEGRATION_STRATEGY

Rigorous Data Analysis

Category E agents ensure methodologically sound analysis across all paradigms.

Explore Category F: Quality