Untitled

The Practice of Statistics for Business and Economics - Chapter 2 Study Notes

Examining Relationships

Scatterplots

Objectives (PSBE Chapter 2.1)

  • Explanatory and Response Variables

  • Scatterplots

  • Interpreting Scatterplots

  • Outliers

  • The Log Transformation

  • Categorical Variables in Scatterplots

Overview of Examining Relationships

  • Most statistical studies involve more than one variable. Key questions include:

    • How was the data collected?

    • Who collected the data?

    • What are the variables of interest, and how are they measured?

    • Are the variables quantitative or qualitative?

    • Do some variables explain or even cause changes in others?

Explanatory and Response Variables

  • Response Variable: A variable that measures or records an outcome of a study; also known as a dependent variable. It is considered the output of interest.

  • Explanatory Variable: A variable that explains changes in the response variable; also known as an independent variable. It is considered the input or predictor variable.

Interpreting Relationships

  1. Start with a graph.

  2. Look for patterns or trends and deviations from the expected behavior.

  3. Utilize statistical models as needed.

Scatterplot

  • A scatterplot visually represents the relationship between two quantitative variables measured on the same individuals.

  • Typically, the explanatory variable is on the x-axis and the response variable is on the y-axis. Each individual is represented as a point.

Scatterplot Example (Education Spending Data)

  • Population and spending data for states:

    • California: Population = 38.7 million, Spending = $110.1 billion

    • Texas: Population = 26.8 million, Spending = $90.5 billion

    • Florida: Population = 19.8 million, Spending = $42 billion

    • New York: Population = 19.7 million, Spending = $74.6 billion

    • [Others follow similarly]

  • The scatterplot helps analyze how one variable affects the other.

Relationships Without Clear Variables

  • Not all plots have distinct explanatory and response variables.

    • Example inquiries:

    • Do calories explain sodium amounts?

    • Does percent return on Treasury bills explain percent return on common stocks?

Interpreting Scatterplots

Relationship Assessment

  • After plotting variables:

    • Form: Identify whether the relationship is linear, curved, clustered, or without a trend.

    • Direction: Assess if the relationship is positive (both variables increase together), negative (one increases while the other decreases), or has no direction.

    • Strength: Evaluate how closely the points align to the identified form (strong vs. weak relationships).

Outliers

  • An outlier in a scatterplot is a data point that deviates significantly from the overall pattern of the relationship.

Form and Direction of the Association

Categories of Relationship Forms

  1. Linear Relationship: Points approximate a straight line.

  2. Nonlinear Relationship: Points follow a curve.

  3. No Relationship: Variability in X does not indicate variability in Y.

Positive and Negative Associations

  • Positive Association: Higher values of one variable correspond with higher values of the other.

  • Negative Association: Higher values of one variable correspond with lower values of the other.

  • No Relationship: X and Y do not correlate; knowledge of one provides no information about the other.

Strength of the Association

  • Strength varies based on how much scatter there is around the regression line:

    • Strong Relationship: One can predict Y reasonably well for a given X.

    • Weak Relationship: Predicted Y values can vary widely for a given X.

Outliers & Influential Observations

  • Outlier: An observation that significantly deviates from the overall data.

  • Influential Observation: An observation upon removal drastically affects the regression results.

The Log Transformation

  • Employed when data is heavily skewed towards larger values. A log transformation can help normalize the data distribution.

Categorical Variables in Scatterplots

Need for Grouping in Analysis

  • Often data must be categorized:

    • Example: Comparison of CO2 emissions by fuel type illustrates trend visibility when colors differentiate categories.

    • Observed trends in men’s vs women’s racing records over time show apparent relationships only when genders are categorized.

When Explanatory Variable is Categorical

  • Scatterplot not feasible; use side-by-side graphs (box plots, mean ± standard deviation) to visualize comparisons across categorical variables.

Correlation

Objectives for Chapter 2.2

  • The Correlation Coefficient (r)

  • Properties of r:

    • Does not distinguish between X and Y.

    • Unitless measure ranging from -1 to 1.

Stronger Association Analysis

  • Two scatterplots with the same data can appear varies based on space surrounding the points; assessment requires quantification.

Correlation Coefficient (r)

  • Measures direction and strength of linear relationships among quantitative variables:

    • Formula: Standard correlation calculations are typically done via software.

  • Properties of r:

    • Ignores response/explanatory distinctions.

    • Not sensitive to changes in measurement units.

    • Positive r indicates direct association; negative indicates inverse association.

    • Strongly affected by outliers.

Range of r

  • Ranges from -1 (perfect negative correlation) to +1 (perfect positive correlation), where 0 indicates no correlation.

Limitations of Correlation

  • Cannot effectively describe nonlinear relationships despite possible strength; maintaining graphical clarity is key.

Residuals and Regression Analysis

Overview

  • The regression line illustrates how Y changes with X, predicting Y for given X values, taking into account previously established distinctions between response and explanatory variables.

Residuals

  • Calculated as the distance between observed Y values and predicted Y values from the regression line, giving qualitative insight into data consistency and model fit.

Key Residual Information

  • Residual plots should ideally display randomness around 0 to imply model adequacy.

Outliers and Influential Points

  • Clear understanding of outliers and their influences on regression is necessary for reliable analysis.

Caution in Correlation/Regression Analysis

Important Considerations

  1. Beware of lurking variables.

  2. Eschew extrapolation beyond the data range.

  3. Interpret correlations with acknowledgment of averages, recognizing data spread can conceal true relationships.

  4. Understand that correlation does not equate causation.