Correlations Is there a correlation between spending on athletics and wins? One would think so, but check out this outlier: College Basketball Budgets (2007) â€“ 339 DI Teams 1. Kentucky -- $9,204,755 339. VMI -- $99,285 VMI won the game 111-103. Up this weekend: 112. Delaware State -- $1,796,416
Homework Assignment Note If you are using the SPSS Student Edition you cannot open datasets with more than 50 variables. Therefore, you need to use: NES2004A_Student.sav Let me know which file you used on the homework.
Homework Assignment Note The variable names are virtually the same: Ideology Measure: v_140a is libcon7 Party ID: partyid3 is the same 2004 Vote: who_2004 is who04_2
Statistical Relationships Andrew Martin PS 372 University of Kentucky
Statistical Relationships Generally speaking, a statistical relationship between two variables exists if the values of the observations for one variable are associated with the observations for the other variable.
Statistical Relationships Knowing two variables are related allows political scientists to make predictions.
How strong is the relationship?
What is the direction or shape of the relationship?
Is it a causal one?
Does it change or disappear if other variables are considered?
Can we conclude the relationship holds for the population?
The level of measurement.
The form of the relationship.
The strength of the relationship.
Numerical summaries of relationships.
Levels of Measurement
Types of Relationships General association. Exists when the values of one variable, X, tend to be associated with specific values of the other variable.
Monotonic Relationships Positive monotonic correlation. High values of one variable (X) are associated with high values of another (Y), and conversely, low values (X) are associated with low values (Y). Negative monotonic correlation. High values of X are associated with low values of Y; low values of X are associated with high values of Y.
Monotonic Relationships In a positive monotonic relationship, the data curve never goes down once on its way up. In a negative monotonic relationship, the data curve never goes up once on its way down.
Positive linear correlation. This type of correlation is a particular type of monotonic relationship in which plotted X-Y values fall on (or at least close to) a straight line. The line slopes upward from left to right.
Negative linear correlation. In this type of correlation, the plotted values of X and Y fall on a straight line that slopes downward from left to right.
Perhaps you remember the following from high school algebra:
Y = mX +b,
Y = Y value
X= X value
m = slope
b = intercept Note: The textbook uses b for the slope coefficient instead of m , and a instead of b for the intercept.
Relationships may have other forms, as when values of X and Y increase together until some threshold is met when they decline.
These are known as curvilinear relationships, and will not be addresses in this class.
Strength of Relationship
Strength of relationship is an indication of how consistently the values of a dependent variable are associated with the values of an independent variable.
Strength of Relationship
12-2a. The values of X and Y are tied together tightly. You could imagine a straight line passing through or very near most points.
Strength of Relationship
12-2b The values tend to be associated â€“ as X increases, so does Y â€“ but the connection is rather weak.
Measures of Association Measures of association are statistics that summarize the relationships between two variables. These measures are typically used to support theoretical or policy claims.
Measures of Association However, a note of caution: These coefficients (1) assume a particular level of measurement â€“ nominal, ordinal, interval and ratio (2) rest on a specific conception of association To interpret its numerical value one has to grasp the kind of association being measured.
Important Properties of Coefficients (1) Null value: Zero typically indicates no association, but there are exceptions. (Ex: Difference of the means) (2) Maximum values: Some coefficients have a maximum values. Many are bounded, with the typical lower bound being 0 and the upper bound being 1. (Ex: Correlation)
Important Properties of Coefficients (3) Strength of the relationship. Subject to lower and upper boundaries, a coefficient's absolute numerical value increases with the strength of the association. (Ex: Regression coefficient) (4) Level of measurement. Nominal, ordinal and quantitative variables require their own type of coefficient. (Ex: Stats for quantitative data)
Important Properties of Coefficients (5) Symmetry. A symmetric measure keeps the same value no matter which variable is treated as dependent or independent. With an asymmetric measures. The coefficient calculated with Y as dependent variable may be differ from the same indicator using X as the dependent variable. (Ex: Correlations)
Important Properties of Coefficients Standardized vs. Unstandardized The measurement scale affects the numerical value of most coefficients of association. Sometimes statisticians transform variables into standardized coefficients so that they all have variances of 1. (Ex: Standard errors)