科研 Skill

定性比较分析QCA

alterlab-qca

Runs crisp-, multi-value, and fuzzy-set Qualitative Comparative Analysis via R’s QCA package. Use for configurational or necessary/sufficient-condition research.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

QCA — Set-Theoretic, Configurational, Equifinal

Skill type: ANALYSIS MODULE. Qualitative Comparative Analysis (Ragin) treats cases as configurations of conditions and asks which combinations are consistently sufficient (or necessary) for an outcome. It is not net-effects regression: it assumes equifinality (several paths to the same outcome), conjunctural causation (conditions work in combination), and asymmetry (the path to Y is not the mirror of the path to not-Y).

Core Mission

QCA IS ABOUT COMBINATIONS AND SET RELATIONS, NOT AVERAGE MARGINAL EFFECTS. CALIBRATE, THEN MINIMIZE.

When to Use This Skill

  • "Run a fuzzy-set QCA / fsQCA on my configurational data."
  • "Which combinations of conditions are sufficient for the outcome?"
  • "Calibrate these variables into set membership and build a truth table."
  • "Test necessary and sufficient conditions across my ~20 cases."

Does NOT Trigger

The request is really about…Route toWhy not this skill
The net effect of one treatment (DiD/IV/RDD/regression)alterlab-causal-inferenceCorrelational/marginal effects, not set relations.
Whether a configurational design is even appropriatealterlab-ssci-design-gateDesign routing, upstream.
Interpretive/thematic analysis of qualitative dataalterlab-qualitative-methodsNot set-theoretic.
Plain descriptive statisticsalterlab-statistical-analysisNo calibration/minimization.

Hand-off rule. When the request matches a row above, open your answer by naming that skill by its exact id (for example alterlab-causal-inference) and saying in one line why it fits better, then stop applying this skill's procedure. Name it even when you also need to ask for missing input. If only part of the request is out of scope, name the skill for that part and do only the in-scope part. A brief general pointer is fine; the named skill does the work.

The R path (verified — QCA package, no Python equivalent)

There is no maintained Python QCA library (the only PyPI option, scpQCA, is stale and non-standard). The field standard is Adrian Dușa's R QCA package. Shell to it via Rscript.

Getting R ready. Check R with Rscript --version (not which). If QCA is missing, install it into a user library together with admisc. A system R often carries an old admisc that blocks current QCA ("admisc 0.34 was found, but > 0.39 is required"):

Rscript -e 'lib <- path.expand("~/R/qca-lib"); dir.create(lib, recursive = TRUE, showWarnings = FALSE); .libPaths(c(lib, .libPaths())); install.packages(c("admisc", "QCA"), lib = lib, repos = "https://cloud.r-project.org")'

Start every later script with .libPaths(c(path.expand("~/R/qca-lib"), .libPaths())), and tell the user the one-time install went into ~/R/qca-lib. Checked 2026-10-15 on R 4.3.3: QCA 3.25 with admisc 0.41 installs and minimize() runs.

The three verified calls:

library(QCA)

# 1. Calibrate raw columns into fuzzy (or crisp) set membership.
#    thresholds for the direct method = c(exclusion, crossover, inclusion), ascending.
df$A <- calibrate(raw$A, type = "fuzzy", thresholds = c(4, 6, 8))

# 2. Build the truth table.
#    incl.cut = consistency cutoff; n.cut = min frequency for an observed row.
tt <- truthTable(df, outcome = "OUT", conditions = c("A","B","C"),
                 incl.cut = 0.80, n.cut = 1, show.cases = TRUE, sort.by = "incl")

# 3. Boolean minimization.
#    conservative = no remainders; parsimonious = include "?"; intermediate = supply dir.exp.
sol_c <- minimize(tt, details = TRUE)                       # conservative
sol_p <- minimize(tt, include = "?", details = TRUE)        # parsimonious
sol_i <- minimize(tt, include = "?", dir.exp = "A, B, C", details = TRUE)  # intermediate
#    dir.exp is an expression of expected conditions: "A, ~B" = A present, B absent expected.

Fit parameters printed: inclS (sufficiency consistency), PRI, covS (raw coverage), covU (unique coverage); necessity uses inclN, RoN. pof() tests parameters of fit for a specific expression; SetMethods adds necessity/sufficiency XY plots, robustness, and clustered-data diagnostics. Full runnable template and the calibration guidance: references/qca_in_r.md.

The discipline QCA demands

  1. Calibration is a theoretical act, not a mechanical one. Anchor the three thresholds (full-out, crossover, full-in) on substantive knowledge; avoid membership exactly 0.5 (it is undefined for logical operations). Justify each anchor.
  2. Report both the parsimonious and intermediate solutions, and be explicit about which logical remainders (counterfactuals) the intermediate solution used — that is where analytic choices hide.
  3. Consistency before coverage. A solution term needs high sufficiency consistency (inclS, commonly ≥ .80) before its coverage is worth interpreting; report PRI to catch simultaneous subset relations in Y and not-Y.
  4. Necessity and sufficiency are separate analyses — test necessary conditions before minimizing for sufficiency; do not read a sufficiency solution as a necessity claim.

Output Template

CALIBRATION:  <per condition: anchors (excl, crossover, incl) + justification>
TRUTH TABLE:  <rows, incl.cut, n.cut, contradictions handled>
SOLUTIONS:    conservative / parsimonious / intermediate (state remainders used)
FIT:          per term inclS, PRI, covS, covU; overall solution consistency + coverage
CLAIM:        configurational (combinations X are sufficient for Y); NOT "X raises Y on average"

References

  • references/qca_in_r.md — full runnable Rscript template, calibration guidance, fit-parameter glossary, and the graceful-degradation note when R is unavailable.

Part of the AlterLab Academic Skills suite.

示例提示词

我想用模糊集定性比较分析找出促成政策成功的组态条件。

所属插件

社会学、政治学与城市研究面向社会科学研究:问卷与访谈设计、质性编码、文本即数据、社会网络、ABM、比较政治与选举数据、RCT 与 DID、城市与区域分析。

安装该插件会一并获得同领域的 58 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

定性比较分析QCA能做什么?

Runs crisp-, multi-value, and fuzzy-set Qualitative Comparative Analysis via R’s QCA package. Use for configurational or necessary/sufficient-condition research。

如何安装定性比较分析QCA?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「定性比较分析QCA」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

定性比较分析QCA需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

定性比较分析QCA经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。