Scatter Plots Mini Lecture Notes

Scatter Plots: Overview

  • Scatter plots provide a graphical way to visualize the association between two variables.

  • When correlations are considered, the explanatory (independent) variable is placed on the x-axis (horizontal axis) and the response (dependent) variable on the y-axis (vertical axis).

  • Each participant is represented by a point with coordinates corresponding to their scores on the two variables; for example, in the dataset about hazardous drinking behavior and stress levels, stress is the explanatory variable on the x-axis and hazardous drinking is the response variable on the y-axis.

  • Example data points mentioned in the transcript:

    • Participant 1: stress = 9, hazardous drinking = 12

    • Participant 2: stress = 2, hazardous drinking = 4

    • A total of 81 participants plotted on the scatter plot

Axes, Points, and Interpretation

  • Center representation: each individual is a dot whose position reflects their scores on the two variables.

  • The scatter plot is a snapshot of the joint distribution of the two variables.

  • The main tasks are to visually inspect the plot for patterns and deviations.

Features to Look For in a Scatter Plot

  • Form: Are there clusters or groupings in the data? Do the points tend to align along a pattern or diagonal? In the example, there are no clear breakaway clusters.

  • Location with respect to axes: Do most data points lie toward a particular region (e.g., lower-left corner indicates generally low stress and low hazardous drinking)?

  • Direction (trend): Is there a diagonal pattern suggesting a relationship between the variables?

  • Strength: How tightly do the points cluster around a pattern or line?

  • Deviations/outliers: Are there points that deviate substantially from the general pattern?

  • Scaling effects: The appearance of strength can be affected by how the axes are scaled.

Example Observations from the Transcript

  • Form: Most data are down toward the lower-left corner (low stress and low hazardous drinking) with no clear cluster formation.

  • Direction: The data appear to trend upwards along a diagonal, suggesting a positive association.

  • Scaling effect: It’s noted that changing the x and y axis scales can make strength appear stronger or weaker.

  • Deviations: Two data points have high hazardous drinking scores but not very high stress, which are deviations from the general pattern.

  • Visual illusion: A red horizontal line may appear angled due to the plotting, reminding us to verify what is truly horizontal.

Types of Associations

  • Negative association: a trend from the top-left to the bottom-right of the plot. Higher-than-average values of one variable tend to be associated with lower-than-average values of the other variable.

  • Positive association: a trend from the bottom-left to the top-right of the plot. Higher-than-average values on one variable tend to be associated with higher-than-average values on the other variable.

  • Weak strength: Data are broadly scattered around the plot, with a loose directional tendency.

  • Moderate strength: Data are more tightly clustered around a line.

  • Strong strength: A clear direction exists and the data are tightly clustered around an imaginary line.

  • Important note: Strength can be influenced by the scale of the axes; extending the range can make the relationship appear stronger.

Specific Deviations/Outliers Discussed

  • Deviations to consider:

    • Two individuals with high hazardous drinking but not correspondingly high stress.

    • An example of an outlier where one individual smokes many cigarettes but has a relatively high life expectancy compared to others with higher smoking levels and lower life expectancy.

Line of Best Fit (Regression Line)

  • Scatter plots can be examined further by adding a line of best fit, which imposes a mathematical model on the data to describe direction, strength, form, and deviations.

  • The line is called the linear regression line or the least squares regression line:

    • Mathematical description: y = \beta0 + \beta1 x</p></li><li><p>Coefficients(sample−based):</p></li><li><p></p></li><li><p>Coefficients (sample-based):</p></li><li><p>\beta1 = \frac{S{xy}}{S_{xx}}</p></li><li><p></p></li><li><p>\beta0 = \bar{y} - \beta1 \bar{x}</p></li></ul></li><li><p>Predictiveuse:Thelinecanbeusedforprediction.Forexample,ifthestresslevelx=17,thepredictedhazardousdrinkingscoreisapproximatelyy^=13,i.e.</p></li></ul></li><li><p>Predictive use: The line can be used for prediction. For example, if the stress level x = 17, the predicted hazardous drinking score is approximately ŷ = 13, i.e.\hat{y} = 13 \quad \text{when} \quad x = 17.(Thisisaone−waypredictionfromxtoy;predictionsarenotguaranteedforallxvalues.)</p></li><li><p>Visualillusion:Thehorizontalredlinemayappearangledduetoplotting,butitcanbecheckedtobehorizontalifadjustedproperly.</p></li><li><p>Extensions:Theregressionlinecanbeextendedtoincludemorevariablestoimprovepredictivepower,whichleadstomultipleregressioninmoreadvancedcourses.</p></li></ul><h3collapsed="false"seolevelmigrated="true">PracticalTakeaways</h3><ul><li><p>Scatterplotsvisualizedataandhelpassessform,direction,andstrengthofassociations.</p></li><li><p>Theyprovideaquickwaytodetectobviousdeviationsfromthepatternandpotentialoutliers.</p></li><li><p>Alineofbestfit(linearregression)addsapredictivecomponent,describingtherelationshipwithasimplemathematicalmodel.</p></li><li><p>Predictionsfromtheregressionlinearedirectional(fromxtoy)anddependontheappropriatenessofthelinearmodelforthedata.</p></li><li><p>Beawarethataxisscalingcaninfluencetheperceivedstrengthoftherelationship.</p></li></ul><h3collapsed="false"seolevelmigrated="true">ConnectionstoEarlierContentandReal−WorldRelevance</h3><ul><li><p>Thisminilecturebuildsontheideathatcorrelationimpliesarelationshipandthatscatterplotsarethefirstgraphicaltooltoexplorethisrelationshipvisuallybeforequantifyingit.</p></li><li><p>Inreal−worlddataanalysis,scatterplotsareusedtoexplorerelationshipsbetweenanytwoquantitativevariables(e.g.,testscoresvs.studyhours,bloodpressurevs.age).</p></li><li><p>Understandingscatterplotslaysthegroundworkforregressionanalysis,modelfitting,andpredictiveanalyticsinlatercourses.</p></li></ul><h3collapsed="false"seolevelmigrated="true">QuickRecapofKeyPoints</h3><ul><li><p>Scatterplotsplacetheexplanatoryvariableonthex−axisandtheresponsevariableonthey−axis;eachpointrepresentsanindividual.</p></li><li><p>Lookforform(clusters),direction(positive/negativeassociation),andstrength(weak/moderate/strong).</p></li><li><p>Scalingcanaffectperceivedstrength;bemindfulofaxisranges.</p></li><li><p>Deviations/outliersmayindicateexceptionstothegeneralpattern.</p></li><li><p>Alineofbestfit(leastsquaresregressionline)providesasimplepredictivemodel:(This is a one-way prediction from x to y; predictions are not guaranteed for all x values.)</p></li><li><p>Visual illusion: The horizontal red line may appear angled due to plotting, but it can be checked to be horizontal if adjusted properly.</p></li><li><p>Extensions: The regression line can be extended to include more variables to improve predictive power, which leads to multiple regression in more advanced courses.</p></li></ul><h3 collapsed="false" seolevelmigrated="true">Practical Takeaways</h3><ul><li><p>Scatter plots visualize data and help assess form, direction, and strength of associations.</p></li><li><p>They provide a quick way to detect obvious deviations from the pattern and potential outliers.</p></li><li><p>A line of best fit (linear regression) adds a predictive component, describing the relationship with a simple mathematical model.</p></li><li><p>Predictions from the regression line are directional (from x to y) and depend on the appropriateness of the linear model for the data.</p></li><li><p>Be aware that axis scaling can influence the perceived strength of the relationship.</p></li></ul><h3 collapsed="false" seolevelmigrated="true">Connections to Earlier Content and Real-World Relevance</h3><ul><li><p>This mini lecture builds on the idea that correlation implies a relationship and that scatter plots are the first graphical tool to explore this relationship visually before quantifying it.</p></li><li><p>In real-world data analysis, scatter plots are used to explore relationships between any two quantitative variables (e.g., test scores vs. study hours, blood pressure vs. age).</p></li><li><p>Understanding scatter plots lays the groundwork for regression analysis, model fitting, and predictive analytics in later courses.</p></li></ul><h3 collapsed="false" seolevelmigrated="true">Quick Recap of Key Points</h3><ul><li><p>Scatter plots place the explanatory variable on the x-axis and the response variable on the y-axis; each point represents an individual.</p></li><li><p>Look for form (clusters), direction (positive/negative association), and strength (weak/moderate/strong).</p></li><li><p>Scaling can affect perceived strength; be mindful of axis ranges.</p></li><li><p>Deviations/outliers may indicate exceptions to the general pattern.</p></li><li><p>A line of best fit (least squares regression line) provides a simple predictive model:\hat{y} = \beta0 + \beta1 x\r

    • Example: with x = 17, predicted y ≈ 13: y^=13whenx=17.\hat{y} = 13 \quad \text{when} \quad x = 17.

    • Extensions to multiple predictors are possible and are covered in more advanced coursework.

    Note: The transcript includes a stray line, "3,003 psi," which does not pertain to the scatter plot content and can be disregarded in the context of the scatter plot discussion.