一键在 Manus 中运行任何 Skill

alterlab-survey-analysis

星标46

分支6

更新时间2026年7月2日 15:44

Analyzes complex-sample survey data with design-based inference — declares a survey design (weights, strata, PSUs/clusters, FPC) before estimating means, totals, proportions, ratios, and quantiles, computes design-adjusted standard errors via Taylor linearization or replicate weights (BRR, Jackknife, Bootstrap), calibrates with post-stratification / raking / GREG, and fits design-adjusted GLMs (linear, logistic, Poisson). Uses samplics (stable Python), the emerging svy successor, or the field-standard R survey + srvyr via Rscript. Use when analyzing GSS/ANES/ESS/DHS/Eurobarometer or any weighted/stratified/clustered survey, when a dataset ships survey weights, or when someone quotes unweighted percentages from a complex survey. For questionnaire and sampling-plan DESIGN prefer alterlab-survey-design; for the sampling-adequacy gate prefer alterlab-ssci-sampling-gate; for causal identification prefer alterlab-causal-inference. Part of the AlterLab Academic Skills suite.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

AlterLab-IEU

AlterLab-IEU/AlterLab-Academic-Skills

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

相关职业SOC

基于 SOC 职业分类

社会科学研究助理生命、物理与社会科学类职业·SOC 19-4061

文件资源管理器

4 个文件

SKILL.md

readonly

同仓库更多 Skills

同仓库

alterlab-skill-finder

AlterLab-IEU/AlterLab-Academic-Skills

The AlterLab front door and multi-agent launcher — routes a task to the right AlterLab skill(s) when the user invokes the suite without naming one, and for a multi-stage goal (or on the keyword 'alterflow', aliases 'alterresearch' / 'ultralab') it CLARIFIES the goal with a few questions, SELECTS the skills the task needs, and runs a dynamic multi-agent workflow composing them (via alterlab-workflow-orchestration, alterlab-research-pipeline, or alterlab-ssci-orchestrator). Triggers on 'use AlterLab skills', 'which AlterLab skill for X', 'is there an AlterLab skill for…', a multi-stage research goal, 'alterflow …', or any generic AlterLab request where the user does not know skill names. It always asks clarifying questions before executing a multi-step run. Use when someone references AlterLab generically, describes a multi-stage goal, or fires the alterflow keyword; when the user already names a specific skill, defer to that skill directly. Part of the AlterLab Academic Skills suite.

2026-07-0246

alterlab-meta-analysis

AlterLab-IEU/AlterLab-Academic-Skills

Runs quantitative meta-analysis — computes effect sizes (Hedges' g / standardized mean difference, log odds/risk ratios) with their variances, pools them under fixed-effect and random-effects models, quantifies heterogeneity (I-squared, tau-squared, Cochran's Q), draws forest and funnel plots, and tests publication bias (Egger's regression, trim-and-fill) — using statsmodels.stats.meta_analysis in Python or the field-standard R metafor via Rscript. It enforces PRISMA reporting and the random- vs fixed-effect decision. Use when pooling effect sizes across studies, running a systematic review's quantitative synthesis, or assessing heterogeneity and publication bias. For finding and screening the literature prefer alterlab-deep-research; for a single study's statistics prefer alterlab-statistical-analysis. Part of the AlterLab Academic Skills suite.

2026-07-0246

alterlab-missing-data

AlterLab-IEU/AlterLab-Academic-Skills

Handles missing data with principled methods — forces an explicit MCAR / MAR / MNAR mechanism statement, then applies multiple imputation by chained equations (MICE) with Rubin's-rules pooling of estimates and standard errors, or full-information maximum likelihood (FIML) where a likelihood/SEM model applies. Uses statsmodels MICE / MICEData in Python or the field-standard R mice via Rscript, and warns that single (mean/regression) imputation and scikit-learn's IterativeImputer return one completed dataset without Rubin's-rules pooling, so they understate standard errors if used as multiple imputation. Use when a dataset has missing values, when choosing an imputation strategy, or when reporting how missingness was handled. For general modeling on complete data prefer alterlab-statistical-analysis; for latent-variable models with FIML prefer alterlab-sem-psychometrics. Part of the AlterLab Academic Skills suite.

2026-07-0246

alterlab-multilevel-models

AlterLab-IEU/AlterLab-Academic-Skills

Fits and reports mixed-effects / multilevel / hierarchical models for clustered, nested, longitudinal, and repeated-measures data — random intercepts and slopes, variance components and the ICC, cross-level interactions, and GLMMs (logistic/Poisson) — using statsmodels MixedLM and bambi (Bayesian on PyMC) in Python, or the field-standard R lme4 / glmmTMB / brms via Rscript. It enforces the reporting items reviews find under-reported: full fixed + random specification, centering, variance components + ICC, estimation method, assumption checks, model comparisons, and effect sizes. Use when data are grouped/nested (students in schools, repeated measures, panel/longitudinal) and the question concerns within- vs between-cluster variation. For general single-level regression prefer alterlab-statsmodels; for panel fixed effects used for causal identification prefer alterlab-causal-inference. Part of the AlterLab Academic Skills suite.

2026-07-0246

alterlab-qualitative-analysis

AlterLab-IEU/AlterLab-Academic-Skills

Analyzes qualitative data as a dispatched pipeline module — codebook development, thematic / framework / content analysis, and inter-coder reliability computed correctly (Krippendorff's alpha as primary via the krippendorff package or a bundled stdlib nominal calculator with bootstrap CIs; Cohen's / Fleiss' kappa via statsmodels) with 95% CIs and thresholds (alpha >= .80 reliable, .667-.80 tentative). It BRANCHES by design: coefficient-based ICR for codebook / content-analytic coding, versus consensus-and-reflexivity for reflexive thematic analysis where a statistic is not the right criterion. Supports human-vs-LLM double-coding with an alpha check against a human gold standard. Use when coding interviews or open-ended text, building a codebook, or reporting intercoder reliability. For topic modeling / embeddings / supervised text classification prefer alterlab-text-as-data; for the reflexivity gate prefer alterlab-ssci-reflexivity-gate. Part of the AlterLab Academic Skills suite.

2026-07-0246

alterlab-ssci-orchestrator

AlterLab-IEU/AlterLab-Academic-Skills

Coordinates a social-science study through the stage-gated methods pipeline — question then design-gate then measurement-gate then sampling-gate then the right analysis module then inference-gate then reporting — holding a single YAML Design Passport that each gate reads and appends, and enforcing gate order with PASS / WARN / BLOCK semantics (design-gate and inference-gate are fail-closed). It is thin: it routes and holds the artifact, it does not run analysis itself. Use when a researcher describes a whole social-science study end to end, asks to run the methods pipeline or a full methodology review, or wants the design-to-inference workflow coordinated rather than a single step. For a single stage trigger that gate directly (alterlab-ssci-design-gate, -measurement-gate, -sampling-gate, -inference-gate); for multi-agent execution mechanics see alterlab-workflow-orchestration. Part of the AlterLab Academic Skills suite.

2026-07-0246

name	alterlab-survey-analysis
description	Analyzes complex-sample survey data with design-based inference — declares a survey design (weights, strata, PSUs/clusters, FPC) before estimating means, totals, proportions, ratios, and quantiles, computes design-adjusted standard errors via Taylor linearization or replicate weights (BRR, Jackknife, Bootstrap), calibrates with post-stratification / raking / GREG, and fits design-adjusted GLMs (linear, logistic, Poisson). Uses samplics (stable Python), the emerging svy successor, or the field-standard R survey + srvyr via Rscript. Use when analyzing GSS/ANES/ESS/DHS/Eurobarometer or any weighted/stratified/clustered survey, when a dataset ships survey weights, or when someone quotes unweighted percentages from a complex survey. For questionnaire and sampling-plan DESIGN prefer alterlab-survey-design; for the sampling-adequacy gate prefer alterlab-ssci-sampling-gate; for causal identification prefer alterlab-causal-inference. Part of the AlterLab Academic Skills suite.
license	MIT
allowed-tools	Read Bash(python:*)
compatibility	Requires (declare in-session, no runtime install on Anthropic API): Python samplics>=0.6 (stable; TaylorEstimator) — or svy>=0.18 (samplics' successor; API still maturing, pin + re-verify) — OR the field-standard R survey>=4.5 + srvyr>=1.3 via Rscript (csSampling+brms for Bayesian design-based models). Runs locally via `uv run python` / `Rscript`; no API key.
metadata	{"skill-author":"AlterLab","version":"1.0.0","depends_on":"alterlab-survey-design (item/sample design), alterlab-ssci-sampling-gate (frame/size gate), alterlab-statistical-analysis; audited by alterlab-ssci-inference-gate"}

Survey Analysis — Declare the Design Before You Estimate Anything

Skill type: ANALYSIS MODULE. Complex-sample surveys (GSS, ANES, ESS, DHS, Eurobarometer) are drawn with stratification, clustering, and unequal probabilities. Analyzing them as if they were a simple random sample underestimates standard errors and yields falsely narrow CIs and wrong tests. The discipline is design-based inference: a declared design object comes first, every estimate flows through it.

Core Mission

YOU MUST WEIGHT (AND DECLARE STRATA + PSUs) BEFORE QUOTING ANY NUMBER FROM A COMPLEX SURVEY.

When to Use This Skill

"Give me the weighted % who [X] from ANES/GSS/DHS, with correct standard errors."
"Why are my survey confidence intervals so narrow?" (← design ignored)
"Post-stratify / rake my sample to census margins."
"Fit a logistic regression on this weighted, clustered survey."

Does NOT Trigger

The request is really about…	Route to	Why not this skill
Writing questionnaire items / choosing a sampling frame	`alterlab-survey-design`	Instrument & sampling design, not weighted analysis.
Whether the sample size / frame is adequate	`alterlab-ssci-sampling-gate`	Sampling-adequacy gate, upstream.
Causal identification (DiD/IV/RDD)	`alterlab-causal-inference`	Design-based survey SEs ≠ causal identification.
Plain unweighted descriptive/inferential stats	`alterlab-statistical-analysis`	No survey design to honor.

The design-object-first rule

Declare weights + strata + PSU/cluster + FPC before any estimate:

weights — the inverse-inclusion-probability weight; scales the sample to the population.
strata — variances are computed within each stratum and pooled. Dropping strata leaves point estimates unchanged but inflates SEs (you lose the variance reduction).
PSU / cluster — the unit of randomization. If whole districts were sampled, the district is the PSU; lower units are not independent. Declaring the PSU is what corrects the SE upward for the clustering.
FPC — finite-population correction when the sampling fraction is non-trivial.

Domain (subpopulation) estimation: subset the design object, never filter the data frame first — filtering discards the strata/PSU structure needed for correct domain SEs.

Weight-type discipline

Distinguish design weights (selection probability only) from post-stratification / calibration weights (also correct for sampling error and non-response). One or the other must always be used; report which. "Weight before quoting any percentage."

Variance estimation — support both families

Taylor linearization — the default analytic method.
Replicate weights — BRR, Jackknife (JKn), Bootstrap. Use these when the data provider ships replicate weights (many public files do); do not re-derive a design they already replicated.

Verified calls (pinned)

Python — samplics (stable):

from samplics.estimation import TaylorEstimator
from samplics.utils.types import PopParam
est = TaylorEstimator(PopParam.mean)
est.estimate(y=df["trust"], samp_weight=df["wt"], stratum=df["strata"], psu=df["psu"],
             fpc=df.get("fpc", 1.0), domain=df.get("region"), deff=True, remove_nan=True)

svy (samplics' successor, import svy) mirrors this with svy.Design(...) / svy.Sample(...); its API is still maturing — pin svy>=0.18 and re-verify against the installed package.

R — survey / srvyr (field standard, fully verified):

library(survey)
des <- svydesign(ids = ~psu, strata = ~strata, weights = ~wt, fpc = ~fpc,
                 data = dat, nest = TRUE)
svymean(~trust, des, deff = TRUE)
svyby(~trust, ~region, des, svymean)                 # domain estimation (keeps structure)
svyglm(trust01 ~ age + educ, design = des, family = quasibinomial())
# replicate weights when provided:
rep <- svrepdesign(weights = ~wt, repweights = "wtrep[0-9]+", type = "JKn", data = dat)
# calibration:
des2 <- rake(des, sample.margins = list(~agecat, ~sex),
             population.margins = list(pop.agecat, pop.sex))

Full Taylor-vs-replicate math, calibration (post-stratification / raking / GREG), and the Python caveats vs the canonical R recipes: references/design_and_variance.md, references/python_vs_r.md.

Reporting checklist (put in every survey result)

DESIGN:     weights (type: design | post-strat/calibrated) + strata + PSU + FPC declared
VARIANCE:   Taylor linearization | replicate weights (BRR/JKn/Bootstrap)
N:          unweighted N  vs  weighted population estimate
DEFF:       design effect per key estimate (how much the design inflates variance)
DOMAINS:    subset of the DESIGN object, not a filtered data frame

References

references/design_and_variance.md — Taylor vs replicate variance, calibration math, DEFF, domain estimation.
references/python_vs_r.md — samplics/svy caveats and the canonical R survey/srvyr recipes.

Part of the AlterLab Academic Skills suite.