Linear regression: Statistical method to find the line of best fit that minimizes the sum of squared residuals between data points and the line
- Organize data into x and y values
- Calculate means of x and y values
- Compute slope using least squares formula
- Find y-intercept using point-slope form
- Calculate correlation coefficient
- Use the model for predictions
Hours studied (x): 2, 3, 4, 5, 6
Exam scores (y): 65, 72, 78, 85, 90
Mean of x: \(\bar{x} = \frac{2+3+4+5+6}{5} = \frac{20}{5} = 4\)
Mean of y: \(\bar{y} = \frac{65+72+78+85+90}{5} = \frac{390}{5} = 78\)
\(\sum x_i = 20\), \(\sum y_i = 390\), \(\sum x_i^2 = 4+9+16+25+36 = 90\)
\(\sum x_iy_i = 2×65 + 3×72 + 4×78 + 5×85 + 6×90 = 130 + 216 + 312 + 425 + 540 = 1623\)
Using: \(m = \frac{n\sum x_iy_i - \sum x_i \sum y_i}{n\sum x_i^2 - (\sum x_i)^2}\)
\(m = \frac{5(1623) - (20)(390)}{5(90) - (20)^2} = \frac{8115 - 7800}{450 - 400} = \frac{315}{50} = 6.3\)
Using: \(b = \bar{y} - m\bar{x}\)
\(b = 78 - 6.3(4) = 78 - 25.2 = 52.8\)
\(y = 6.3x + 52.8\)
For 7 hours of study: \(y = 6.3(7) + 52.8 = 44.1 + 52.8 = 96.9\)
Predicted score: approximately 97 points
The regression line is \(y = 6.3x + 52.8\) and a student who studies 7 hours is predicted to score approximately 97 points.
• Least squares regression: Minimizes sum of squared residuals
• Slope formula: \(m = \frac{n\sum xy - \sum x \sum y}{n\sum x^2 - (\sum x)^2}\)
• Y-intercept formula: \(b = \bar{y} - m\bar{x}\)
Quadratic regression: Statistical method to find the parabola that best fits data points, useful for modeling projectile motion
The height increases, reaches a peak, then decreases - characteristic of quadratic motion
Height peaks at (5, 125), suggesting vertex form might be useful
General form: \(y = ax^2 + bx + c\)
Using point (0, 0): \(0 = a(0)^2 + b(0) + c\), so \(c = 0\)
Using (1, 45): \(45 = a(1)^2 + b(1)\), so \(a + b = 45\)
Using (2, 80): \(80 = a(2)^2 + b(2)\), so \(4a + 2b = 80\)
From: \(a + b = 45\) and \(4a + 2b = 80\)
Multiply first equation by 2: \(2a + 2b = 90\)
Subtract from second: \(2a = -10\), so \(a = -5\)
Substitute: \(-5 + b = 45\), so \(b = 50\)
\(y = -5x^2 + 50x\)
For quadratic \(y = ax^2 + bx + c\) with \(a < 0\), vertex is at \(x = -\frac{b}{2a}\)
\(x = -\frac{50}{2(-5)} = -\frac{50}{-10} = 5\)
Maximum height: \(y = -5(5)^2 + 50(5) = -125 + 250 = 125\)
(0,0): \(-5(0)^2 + 50(0) = 0\) ✓
(1,45): \(-5(1)^2 + 50(1) = 45\) ✓
(2,80): \(-5(4) + 50(2) = 80\) ✓
(5,125): \(-5(25) + 50(5) = 125\) ✓
The quadratic model is \(y = -5x^2 + 50x\) and the maximum height of 125 units occurs at 5 seconds.
• Quadratic motion: Parabolic trajectory typical for projectiles
• Vertex formula: \(x = -\frac{b}{2a}\) for maximum when \(a < 0\)
• System of equations: Use known points to solve for coefficients
Exponential regression: Statistical method to find the exponential function that best fits data points with multiplicative growth
Look for constant ratios between consecutive y-values:
150/100 = 1.5, 225/150 = 1.5, 337/225 ≈ 1.5, 506/337 ≈ 1.5
Constant ratio of 1.5 suggests exponential growth
When x = 0, y = 100, so the initial population is 100
The constant ratio of 1.5 is the growth factor
\(y = 100 \cdot (1.5)^x\)
(0,100): \(100 \cdot (1.5)^0 = 100 \cdot 1 = 100\) ✓
(1,150): \(100 \cdot (1.5)^1 = 100 \cdot 1.5 = 150\) ✓
(2,225): \(100 \cdot (1.5)^2 = 100 \cdot 2.25 = 225\) ✓
(3,337): \(100 \cdot (1.5)^3 = 100 \cdot 3.375 = 337.5 \approx 337\) ✓
After 6 hours: \(y = 100 \cdot (1.5)^6\)
\((1.5)^6 = 11.390625\)
\(y = 100 \times 11.390625 = 1,139.0625 \approx 1,139\)
The bacterial population grows exponentially with a 50% hourly growth rate
After 6 hours, the population is predicted to be approximately 1,139 bacteria
The exponential model is \(y = 100 \cdot (1.5)^x\) and the predicted population after 6 hours is approximately 1,139 bacteria.
• Exponential pattern: Constant ratios between consecutive terms
• Growth factor: Ratio of consecutive y-values gives base of exponential
• Initial value: Value when x = 0 gives coefficient
Regression analysis: Statistical method for determining the relationship between variables by finding the best-fitting curve through data points
Correlation coefficient (r): Measures the strength and direction of a linear relationship, ranging from -1 to 1
Coefficient of determination (R²): Proportion of variance in dependent variable explained by independent variable, ranging from 0 to 1
Residual: Difference between observed value and predicted value from the regression model
Least squares method: Technique that minimizes the sum of squared residuals to find the best-fitting line
Linear regression: Fits a straight line to data points, appropriate for constant rate of change
Quadratic regression: Fits a parabolic curve to data, appropriate for curved relationships
Exponential regression: Fits an exponential curve to data, appropriate for multiplicative growth/decay
- Data collection: Gather paired observations (x, y) for analysis
- Data visualization: Create scatter plot to identify potential patterns
- Pattern recognition: Determine if relationship appears linear, quadratic, or exponential
- Model selection: Choose appropriate regression model based on pattern
- Parameter estimation: Calculate model coefficients using regression formulas
- Goodness-of-fit assessment: Calculate R² and examine residuals
- Model validation: Test model against additional data if available
- Prediction: Use validated model for forecasting or interpolation
• Linear regression: \(y = mx + b\) where \(m = \frac{n\sum xy - \sum x \sum y}{n\sum x^2 - (\sum x)^2}\) and \(b = \bar{y} - m\bar{x}\)
• Correlation interpretation: |r| > 0.8 indicates strong correlation, 0.5 < |r| < 0.8 moderate, |r| < 0.5 weak
• Quadratic regression: \(y = ax^2 + bx + c\) fits parabolic patterns
• Exponential regression: \(y = a \cdot b^x\) fits multiplicative growth/decay patterns
• Model comparison: Choose model with highest R² that makes contextual sense
Model comparison: Process of evaluating multiple regression models to determine which best represents the data based on goodness-of-fit measures
y-values: 10, 18, 30, 52, 90
First differences: 8, 12, 22, 38 (not constant) → Not linear
First differences: 8, 12, 22, 38
Second differences: 4, 10, 16 (not constant) → Not perfectly quadratic
Ratios: 18/10=1.8, 30/18≈1.67, 52/30≈1.73, 90/52≈1.73
Ratios are approximately constant → Exponential pattern
Linear: \(y = 20x - 10\), R² ≈ 0.89
Quadratic: \(y = 2x^2 + 4x + 4\), R² ≈ 0.99
Exponential: \(y = 6.5 \cdot (1.7)^x\), R² ≈ 0.98
Quadratic model has R² ≈ 0.99, highest among models
Despite non-constant second differences, quadratic model provides best fit
\(y = 2x^2 + 4x + 4\)
(1,10): \(2(1)^2 + 4(1) + 4 = 10\) ✓
(2,18): \(2(4) + 8 + 4 = 18\) ✓
(3,30): \(2(9) + 12 + 4 = 30\) ✓
The quadratic model \(y = 2x^2 + 4x + 4\) provides the best fit with R² ≈ 0.99
Though ratios are approximately constant, the quadratic model captures the curvature better
The quadratic model \(y = 2x^2 + 4x + 4\) provides the best fit with R² ≈ 0.99, though the data shows mixed characteristics.
• Model comparison: Choose model with highest R² value
• Pattern recognition: Look for constant differences or ratios
• Context consideration: Sometimes real data has mixed characteristics
Business modeling: Application of statistical methods to business data to identify trends and make predictions about future performance
Sales: $50k, $55k, $60k, $67k, $75k
First differences: 5, 5, 7, 8 (not constant)
Second differences: 0, 2, 1 (not constant but showing acceleration)
Linear: \(y = 6.2x + 44.4\), R² ≈ 0.96
Quadratic: \(y = 0.8x^2 + 4.2x + 45\), R² ≈ 0.99
Quadratic model provides better fit
Using the quadratic regression: \(y = 0.8x^2 + 4.2x + 45\)
(1,50): \(0.8(1) + 4.2(1) + 45 = 50\) ✓
(2,55): \(0.8(4) + 4.2(2) + 45 = 55\) ✓
(3,60): \(0.8(9) + 4.2(3) + 45 = 60.6 \approx 60\) ✓
\(y = 0.8(8)^2 + 4.2(8) + 45\)
\(y = 0.8(64) + 33.6 + 45\)
\(y = 51.2 + 33.6 + 45 = 129.8\)
The quadratic model suggests accelerating growth (sales increasing at an increasing rate)
Predicted sales of $129.8k for Month 8 represent continued acceleration
Quadratic growth suggests increasing efficiency or market penetration
Management should consider capacity planning for continued growth
The quadratic model \(y = 0.8x^2 + 4.2x + 45\) best fits the sales data, predicting $129.8k in sales for Month 8.
• Business modeling: Choose model that best represents the trend
• Accelerating growth: Quadratic models capture increasing rates of change
• Prediction caution: Consider business sustainability of trends
Correlation coefficient (r): A statistical measure that quantifies the strength and direction of a linear relationship between two variables, ranging from -1 to 1. Values near ±1 indicate strong correlation, while values near 0 indicate weak correlation.
Coefficient of determination (R²): The proportion of the variance in the dependent variable that is predictable from the independent variable(s), expressed as a percentage from 0 to 1. It represents the goodness of fit of the regression model.
Residual: The difference between an observed value and the value predicted by the regression model. Residual = Observed - Predicted. Analyzing residuals helps assess model adequacy.
Least squares regression: A statistical method that finds the line (or curve) that minimizes the sum of the squared residuals, providing the best fit to the data.
Linear regression: A regression model that assumes a linear relationship between variables, taking the form \(y = mx + b\), where m is the slope and b is the y-intercept.
Quadratic regression: A regression model that fits a parabolic curve to data, taking the form \(y = ax^2 + bx + c\), useful for modeling relationships with curvature.
Exponential regression: A regression model that fits an exponential curve to data, taking the form \(y = a \cdot b^x\), useful for modeling multiplicative growth or decay.
Model selection: The process of choosing the most appropriate mathematical model to represent a given dataset based on statistical measures and real-world context.
- Data preparation: Organize data into ordered pairs (x, y) and verify data quality
- Visual inspection: Create scatter plot to identify potential patterns and outliers
- Pattern analysis: Examine first differences, second differences, and ratios to identify model type
- Model selection: Choose between linear, quadratic, or exponential based on pattern analysis
- Parameter estimation: Calculate regression coefficients using appropriate formulas
- Goodness-of-fit assessment: Calculate R², analyze residuals, and validate model assumptions
- Model interpretation: Understand the meaning of parameters in real-world context
- Prediction: Use validated model for interpolation or cautious extrapolation
• Linear model: \(y = mx + b\), constant rate of change, R² indicates linearity strength
• Quadratic model: \(y = ax^2 + bx + c\), changing rate of change, parabolic shape
• Exponential model: \(y = a \cdot b^x\), constant percentage change, rapid growth/decay
• Correlation interpretation: |r| > 0.8 = strong, 0.5 < |r| < 0.8 = moderate, |r| < 0.5 = weak
• Model selection: Choose model with highest R² that makes contextual sense
• Residual analysis: Random residuals indicate good model fit
• Outlier detection: Points significantly distant from regression line may need special attention
\(f_1(x) = 2x + 3\) (linear)
\(f_2(x) = 0.5x^2 + 2x + 1\) (quadratic)
\(f_3(x) = 2 \cdot 1.3^x\) (exponential)
Analysis: The chart shows how different models behave differently for the same x-values.
- \(f_1(x) = 2x + 3\): Linear growth with constant rate
- \(f_2(x) = 0.5x^2 + 2x + 1\): Quadratic growth with increasing rate
- \(f_3(x) = 2 \cdot 1.3^x\): Exponential growth with multiplicative rate