Your client hands you a data dictionary and asks for a "deeper understanding" of one variable at a time. That is univariate exploration, and the graders reward candidates who match the right summary and the right chart to the variable's type.
Split by variable type first. Univariate technique depends entirely on whether the variable is numeric or categorical. Choose the summary and chart from the matching column, and you have the point.
KEY: Univariate looks at one variable in isolation. If your answer mentions a second variable, a relationship, or a model, you have drifted into bivariate or modeling territory and lost the point.
Center: mean versus median. The mean is the arithmetic average; add the values and divide by the count. The median is the middle value when the data is sorted. The mean gets pulled toward extreme values; the median does not. So comparing them diagnoses skew. Mean well above median means a long right tail. Mean below median means a long left tail. Equal means roughly symmetric.
Common mistakes
- Suggesting a bivariate technique. When asked for a univariate method, a scatterplot is wrong because it involves two variables. Graders flagged this as the most common error.
- Recommending a modeling technique. A regression or GLM is not data exploration. Univariate exploration precedes modeling.
- Reaching for a pie chart. A pie chart earns only partial credit; the bar chart is the intended categorical tool.
Bottom line
- Univariate means one variable at a time; never a scatterplot or a modeling technique.
- Numeric variables get mean, median, variance, standard deviation, and percentiles.
- Numeric variables get histograms and boxplots; categorical variables get bar charts of frequency.
- Mean chases the tail; median resists it. A mean far above the median signals right skew.
Exam shortcut
Classify the variable first: numeric gets a histogram or boxplot plus mean, variance, and percentiles; categorical gets a bar chart of frequencies. If your proposed technique names two variables or a model, delete it, the question wanted univariate. When describing a distribution, always add one sentence linking its skew or spike to the modeling consequence, because graders dock answers that stop at shape.
The full lesson (about 1,608 words, 11 min read) adds 2 worked examples, all 6 common mistakes, a self-check, free in the app.
Learning objectives
- 2e
Browse all free Exam PA lessons or jump into free Exam PA practice questions.