DTSC MIDTERM FINAL REVIEW HELP SHEET

Two numerical variables

Scatterplot used to examine the relationship between two numeric variables

Numerical and categorical variables

Side-by-side boxplot used to compare a numeric variable across categories

Two categorical variables

Stacked bar chart or ribbon plot used to show joint distribution

One categorical variable

Barchart used to display counts or proportions

Stacked boxplot

Not a valid or standard plot and typically a trap answer

Scatterplot

Used to look for patterns or trends between two numeric variables

Correlation

Measures the strength and direction of a linear relationship

Linear regression

Models and predicts a numeric response using one or more predictors

Indicator variable

A binary variable coded as 0 or 1 representing group membership

Interaction variable

A variable created by multiplying predictors to test whether effects differ across groups

Supervised classification

Modeling with a known categorical response variable

Supervised regression

Modeling with a known numeric response variable

Unsupervised classification

Grouping or clustering data without a known response variable

Unsupervised regression

Finding structure or patterns in numeric data without labeled outcomes

Consequentialist theories

Evaluate actions based on outcomes and consequences

Deontological theories

Focus on duties rights fairness and justice

Utilitarianism

A form of consequentialism focused on maximizing overall benefit

Virtue theories

Focus on integrity character and moral responsibility

R squared

The proportion of variability in the response explained by the predictors

R squared interpretation

Measures explanatory power not prediction accuracy

Residual

Observed value minus predicted value

Residual interpretation

Positive residual means underprediction negative residual means overprediction

P-value

Probability of observing results as extreme as the data assuming the null hypothesis is true

P-value interpretation

Small p-value indicates statistical significance large p-value indicates insufficient evidence

Mileage_Porsche coefficient interpretation

Tests whether mileage affects price differently for Porsches compared to the baseline group

Non-significant interaction term

Indicates no statistical evidence that effects differ across groups

Modeling as a socio-technical loop

Models influence society and societal decisions influence future data and models

RMSE

Measures the typical size of prediction errors in the same units as the response

RMSE interpretation

Lower RMSE indicates better predictive accuracy

F-test

Evaluates whether a regression model is useful overall

F-test large value interpretation

Large F-statistic with small p-value indicates at least one predictor has a non-zero effect

F-test small value interpretation

Small F-statistic with large p-value indicates the model is not useful overall

Residuals vs fitted plot

A plot used to check whether prediction errors behave randomly

Good residuals vs fitted pattern

Errors appear random and unrelated to predictions indicating assumptions are reasonable

Bad residuals vs fitted pattern

Errors show systematic behavior indicating the model missed structure or violated assumptions

QQ plot

Used to assess whether residuals are approximately normally distributed

Bootstrapping

Resampling with replacement from observed data

Purpose of bootstrapping

Used to estimate uncertainty when distributional assumptions may not hold