education-descriptive-statistics

# Education Descriptive Statistics ## Goal Generate clear descriptive summaries of education research data so the user understands the sample, variables, item distributions, group balance, and baseline patterns before deeper analysis. ## Use After Use after: - `education-quantitative-data-cleaning` - `education-survey-instrument-design` - `education-quantitative-study-design` - `education-program-evaluation` - `education-psychometric-scale-development` - `education-learning-analytics-design` Do not expose the skill name to users. Present it as "描述性统计与样本概览". ## Inputs - Cleaned dataset - Data dictionary - Variable roles: demographic, grouping, scale/item, outcome, timepoint, class/school - Scale score table if available - Planned analyses ## Workflow 1. Identify variable types: - categorical: gender, grade, group, school type - ordinal: Likert items, rating levels - continuous: scores, scale means, time, counts - nested IDs: class, school, teacher - timepoint: pre/post/follow-up 2. Produce sample description: - N - participant categories - school/class/teacher structure - group allocation - attrition if longitudinal/intervention 3. Produce variable descriptives: - mean, SD, median, min, max - frequencies and percentages - missing counts - item-level distributions 4. Produce group/timepoint descriptives: - experimental vs control - pre vs post - grade/class/school groups - teacher/student groups if relevant 5. Generate cross-tabulations: - group x gender - group x grade - group x school/class - intervention completion x group 6. Recommend visualizations. 7. Flag descriptive risks: - very small groups - skewed variables - ceiling/floor effects - unbalanced baseline - suspicious missing patterns 8. Draft descriptive-results wording. ## Tool Calls Dependency policy: inspect the current environment first. Do not install packages, clone repositories, or change the user's environment without explicit approval. Prefer a project-local Python virtual environment or R library, and record package versions used for reproducibility. ### R ```r install.packages(c("tidyverse", "psych", "gtsummary", "gt", "janitor", "skimr")) library(tidyverse) library(psych) library(gtsummary) library(janitor) library(skimr) ``` Overall descriptives: ```r psych::describe(data[, c("score_pre", "score_post", "self_efficacy")]) ``` Frequency table: ```r data |> count(gender) |> mutate(percent = n / sum(n) * 100) ``` Cross-tab: ```r janitor::tabyl(data, group, gender) |> janitor::adorn_percentages("row") ``` Group descriptives: ```r data |> group_by(group) |> summarise( n = n(), mean = mean(score_post, na.rm = TRUE), sd = sd(score_post, na.rm = TRUE) ) ``` Publication-style table: ```r data |> select(group, gender, grade, score_pre, score_post) |> tbl_summary(by = group) ``` ### Python ```bash pip install pandas numpy scipy seaborn matplotlib tabulate ``` Overall descriptives: ```python df[["score_pre", "score_post", "self_efficacy"]].describe() ``` Frequencies: ```python df["gender"].value_counts(dropna=False) df["gender"].value_counts(normalize=True, dropna=False) * 100 ``` Cross-tab: ```python pd.crosstab(df["group"], df["gender"], normalize="index") * 100 ``` Group descriptives: ```python df.groupby("group")["score_post"].agg(["count", "mean", "std", "median", "min", "max"]) ``` ### SPSS ```text Analyze -> Descriptive Statistics -> Frequencies Analyze -> Descriptive Statistics -> Descriptives Analyze -> Descriptive Statistics -> Explore Analyze -> Descriptive Statistics -> Crosstabs Analyze -> Compare Means -> Means Graphs -> Chart Builder ``` ## Output Format ### 1. Sample Description Table | Characteristic | Category | n | % | |---|---|---|---| ### 2. Continuous Variable Descriptives | Variable | N | Mean | SD | Median | Min | Max | Missing | |---|---|---|---|---|---|---|---| ### 3. Group/Timepoint Descriptives | Group | Timepoint | N | Mean | SD | Median | Min | Max | |---|---|---|---|---|---|---|---| ### 4. Cross-Tabulation | Row Variable | Column Variable | Pattern | Note | |---|---|---|---| ### 5. Visualization Plan | Purpose | Recommended Chart | Variables | |---|---|---| | Sample composition | Bar chart | Gender/grade/group | | Score distribution | Histogram/density | Achievement/scale score | | Group comparison | Boxplot/violin plot | Group x outcome | | Pre-post change | Line/slope chart | Timepoint x score | | Relationship | Scatterplot | Two continuous variables | ### 6. Descriptive Findings Draft ```text The final sample included [N] participants from [schools/classes]. [x%] were in the intervention group and [x%] were in the comparison group. The mean pre-test writing score was [M] (SD = [SD]), while the mean post-test score was [M] (SD = [SD]). Descriptive results suggested [brief pattern], which was further examined using [planned inferential analysis]. ``` Use Chinese if the paper is Chinese: ```text 本研究最终纳入 [N] 名学生，来自 [学校/班级数量]。实验组占 [x%]，对照组占 [x%]。前测写作成绩均值为 [M]（SD = [SD]），后测成绩均值为 [M]（SD = [SD]）。描述性结果显示 [简要模式]，后续将通过 [统计方法] 进一步检验。 ``` ## Education-Specific Checks - Always describe grade, class, school, and teacher where relevant. - For interventions, report baseline group balance descriptively before inferential tests. - For longitudinal/pre-post data, report attrition by group. - For scale items, inspect ceiling/floor effects common in student attitude data. - For nested data, report number of schools/classes, not only student N. - For small samples, avoid overinterpreting mean differences. ## Quality Rules - Descriptive statistics do not prove significance or causality. - Report missing values and valid N clearly. - Use SD for approximately continuous variables; use frequencies for categorical variables. - Do not average categorical codes such as gender or school type. - Do not hide highly unbalanced groups. - If Likert items are summarized with means, acknowledge ordinal nature when appropriate; scale scores are usually more defensible. ## User-Facing Closure End by routing to the next likely analysis: ```text 描述性统计已经能说明样本和变量的基本情况。接下来可以进入三条路线：A. 信效度分析，B. t 检验/方差分析/相关回归，C. 可视化和结果表格整理。你希望先看哪一部分？如果你不确定，我会按论文设计自动推荐。 ```