Univariate and Bivariate Statistical Analysis
Levels of Measurement and Variable Types
- Interval Variables: Variables lacking a meaningful absolute zero; a value of 0 represents an actual measured point rather than non-existence (e.g., temperature).
- Ratio Variables: Variables possessing a true zero point where 0 signifies the absence of the attribute (e.g., zero age, zero minutes).
- Analytical Categories:
- Univariate Analysis: Examination of a single variable in isolation (e.g., referendum voting outcomes).
- Bivariate Analysis: Examination of two variables simultaneously to explore relationships (e.g., one predictor and one outcome).
- Multivariate Analysis: Examination involving multiple variables simultaneously.
Univariate Graphical Displays and Data Issues
- Frequency Tables: Used in descriptive statistics to display counts and percentages across categories.
- Missing Data Treatment: Omitting non-responses alters percentage distributions (e.g., among 90 gym respondents, excluding 1 missing response shifts the 21 to 30 age bracket from 43.3% of all respondents to 43.8% of valid responses).
- Data Displays:
- Bar Chart: Displays distinct categories; not used for continuous trends.
- Pie Chart: Displays proportions of a whole.
- Histogram: Displays continuous numerical intervals rather than distinct categories.
Measures of Central Tendency and Dispersion
- Central Tendency:
- Mode: The most frequently occurring value or category in a distribution.
- Median: The middle observation when data are ranked ordinally; resistant to extreme outliers.
- Mean: Arithmetic average computed by dividing the sum of values by the total count; susceptible to skew from extreme values.
- Dispersion:
- Range: Difference between the highest and lowest values (highest−lowest; e.g., 100−40=60); sensitive to outliers.
- Interquartile Range (IQR): Difference between the 75th percentile and the 25th percentile; eliminates extreme values to reflect middle-spread dispersion.
- Standard Deviation (SD): Measure of the average discrepancy between individual data points (X) and the mean; large SD denotes wide variation, while low SD denotes clustering.
Bivariate Analysis, Causation, and Confounds
- Criteria for Causal Relationships:
- Correlation: Statistical association between the two variables.
- Temporal Order: The predictor must chronologically precede the outcome.
- Absence of Confounds: No extraneous variable creates a spurious link between predictor and outcome.
- Confounding Variables: Extraneous factors that simultaneously influence both the predictor and the outcome (e.g., age acting as a confound in an apparent correlation between shoe size and IQ).
Correlation Coefficients and Explained Variance
- Direction and Magnitude: Ranges from −1 (perfect negative relationship) through 0 (no relationship) to +1 (perfect positive relationship).
- Bivariate Association Coefficients:
- Pearson Correlation Coefficient (r): Evaluates linear relationships between two interval or ratio variables (ranges from −1 to +1).
- Kendall toffee index: Measures direct association between ranked ordinal variables, or between an ordinal and an interval/ratio variable.
- Spearman's role coefficient: Measures direct association between ordinal variables (ranges from −1 to +1).
- Krabbe: Measures association between two categorical/nominal variables (ranges from 0 to 1).
- Coefficient of Determination (R2):
- Obtained by squaring the Pearson correlation coefficient (r).
- Quantifies the proportion of variance in the outcome accounted for by the predictor(s) (e.g., r=0.5 yields R2=0.25, meaning 25% of variance is explained).
- A gym model using age and gender to predict minutes spent using weights with an r of 0.478 produces an R2 explaining 22.8% of the variation.
Questions & Discussion
- Definition of Mode:
- Prompt: Inquiry regarding what constitutes the mode.
- Response: The most common number or most frequently occurring value/category in a population.
- Definition of Standard Deviation:
- Prompt: Inquiry whether standard deviation is the spread over the average mean.
- Response: It represents the average discrepancy and distribution of individual data points relative to the mean value.
- Predictors versus Confounds:
- Prompt: Inquiry whether predictors and confounds are identical concepts.
- Response: They are distinct; a confounding variable influences both the predictor and the outcome simultaneously.
- Meaning of Temporal Order:
- Prompt: Clarification on whether temporal order equates to time order.
- Response: Temporal order requires the predictor event to occur prior in time to the observed outcome.
- Negative Association Logic (Domestic Abuse and Mental Health):
- Prompt: Clarification on why decreasing domestic abuse cases increases mental health.
- Response: A reduction in societal domestic abuse cases leads to an increase in general population mental health and safety, forming a negative correlation.
- Function of the Squared Correlation (R2):
- Prompt: Inquiry regarding why r is squared and what weighting represents.
- Response: While r reflects the strength and direction of association, R2 defines the specific weight or proportion of outcome variance directly explained by the included model variables versus unaccounted factors.