菜单
Methods for calculating statistics and estimates using economic data. Use when user asks to calculate or estimate quantities like percentiles, means, medians, shares, counts, or regression coefficients or regression-adjusted statistics.
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。
基于 SOC 职业分类
Provides useful economic data sources and ways to retrieve the data. Use when user asks for "prices", "inflation", "employment", "unemployment", "inequality", "productivity", "poverty", "CPS", "Census", "BLS", "BEA", "FRED", "IPUMS", "SWADL", "State of Working America Data Library".
Techniques for retrieving and investigating remote codebases and their documentation. Use when user provides a GitHub/GitLab URL to look at, asks to follow conventions from a remote repo, asks "how does package/library X implement/handle Y", or references any external repository's code, structure, or patterns. Examples: "follow the conventions in https://github.com/...", "look at how repo X does Y", "check this GitHub link".
Common ways to validate or benchmark data analysis estimates using external sources. Use to confirm or test the plausibility of results from an economic analysis.
| name | calculate-methods |
| description | Methods for calculating statistics and estimates using economic data. Use when user asks to calculate or estimate quantities like percentiles, means, medians, shares, counts, or regression coefficients or regression-adjusted statistics. |
Often data over time is very noisy and moving averages can better extract signals. Use the R package slider for calculations over sliding windows like moving averages.
Use weighted.mean(vector, w = weight) for weighted means.
Use the MetricsWeighted package for weighted medians and percentiles: weighted_median(vector, w = weight) and weighted_quantile(vector, w = weight, probs = c(0.25, 0.50, 0.75)).
When using data that is assumed to be continuous but in practice is "bunched" at discrete points -- like reported hourly wages -- it can be helpful to smooth the estimated percentiles. One technique is to take a weighted average of nearby percentiles. For example, consider using the epidatatools package averaged_median(vector, w = weight) and averaged_quantile(vector, w = weight, probs = c(0.25, 0.50, 0.75)).
