Correlation & Regression - IB Mathematics Applications and Interpretation

A real estate agent uses the linear regression equation $P = 2150A + 45000$ to model the price, $P$ , of a house in US dollars based on its floor area, $A$ , in square metres. Which statement is the correct interpretation of the slope of this regression line?
Answer: For each additional square metre of floor area, the price of the house is expected to increase by ($2150). The slope of a linear regression equation $y = mx + c$ represents the change in the dependent variable ( $y$ ) for a one-unit increase in the independent variable ( $x$ ). In this model, $P = 2150A + 45000$ , the slope is 2150. This means for each one-unit increase in area $A$ (i.e., 1 square metre), the price $P$ is predicted to increase by ($2150). - A is incorrect because it confuses the slope with a minimum price. - B is incorrect because it swaps the dependent and independent variables. - D is incorrect because it describes the y-intercept, not the slope.
Which of the following is a valid conclusion that can be drawn from this information?
Answer: There is a strong linear association where higher hours of television watching correspond to lower fitness scores. Correlation measures the strength and direction of a linear association between two variables, but it does not imply causation. A correlation coefficient of $r = -0.92$ indicates a strong, negative, linear relationship. - C correctly states this association without implying cause and effect. - A and B are incorrect because they infer a causal relationship, which correlation alone cannot establish. - D is incorrect because it makes a definitive prediction of an increase, which is not guaranteed and also implies causation.
How will the addition of this new point affect the regression line?
Answer: The regression line will not change. When you encounter questions about adding points to regression lines, think about how the least squares method works—it finds the line that minimizes the sum of squared residuals (distances from points to the line). If a new point lies directly on the existing regression line, its residual is zero. This means adding it doesn't increase the sum of squared residuals at all. Since the original line already minimized this sum for the original data, and the new point contributes nothing to increase it, the same line continues to be optimal for the expanded dataset. Mathematically, the regression line remains the best fit because no adjustments to slope or y-intercept could reduce the sum of squared residuals further. The new point essentially "agrees" with the existing relationship between variables. Answer choice A is incorrect because there's no mechanism for the slope to increase—the new point doesn't pull the line in any particular direction since it lies exactly on the current line. Answer choice B fails for the same reason regarding slope decrease. Answer choice C suggests the y-intercept changes while slope stays constant, but this is impossible since any change to the y-intercept would move the line away from some of the original points, increasing the sum of squared residuals unnecessarily. The correct answer is D—the regression line will not change. Study tip: Remember that points lying exactly on a regression line have zero residual. Since regression minimizes residuals, adding zero-residual points never changes the optimal line. Watch for this principle in other regression scenarios on the exam.
What is the most likely conclusion an analyst should draw from this residual plot?
Answer: The relationship between the variables is non-linear. When you encounter residual plots in regression analysis, you're examining whether your model captures the true relationship in the data. Residuals are the differences between observed and predicted values, and their pattern reveals crucial information about model adequacy. A parabolic curve in the residual plot indicates that your linear model is systematically missing something important. When residuals follow a curved pattern, it means the relationship between your variables isn't actually linear—it has curvature that a straight line cannot capture. This systematic pattern suggests you need a quadratic or other non-linear model instead. Option D correctly identifies this non-linear relationship. The curved residual pattern is a classic diagnostic sign that the true relationship has curvature. Option A is wrong because a good linear fit would produce residuals scattered randomly around zero with no discernible pattern. A parabolic curve indicates poor fit. Option B misses the point entirely. Correlation strength isn't what residual plots assess—they evaluate whether your chosen model type (linear vs. non-linear) is appropriate. You could have strong correlation but still need a non-linear model. Option C describes heteroscedasticity (changing variance), which would appear as a fan or funnel shape in residuals, not a parabolic curve. The curve pattern specifically indicates wrong model form, not changing variance. Remember this key principle: random scatter in residuals = good model fit, while systematic patterns (curves, funnels, waves) = model problems. Curved residuals almost always mean you need a non-linear approach.
Which statement provides the best interpretation of this $r^2$ value?
Answer: 55% of the variation in monthly sales can be explained by the variation in advertising spending. The coefficient of determination, $r^2$ , measures the proportion of the variance in the dependent variable that is predictable from the independent variable(s). It is expressed as a percentage. - C gives the correct statistical interpretation. - A is incorrect. $r^2$ does not describe how many points are on the line itself. - B is incorrect because $r^2$ is not a probability of a specific outcome. - D is incorrect because it confuses the coefficient of determination ( $r^2$ ) with the correlation coefficient ( $r$ ).
A researcher finds a strong correlation of $r=0.95$ between a country's annual chocolate consumption per capita and the number of Nobel laureates it has per 10 million people. Which of the following is the most sound conclusion?
Answer: This high correlation is likely coincidental or due to a lurking variable, such as national wealth, affecting both. When you encounter correlation problems on the IB exam, remember that correlation never implies causation – even when the correlation coefficient is very high like $r = 0.95$ . The correct answer is D because a strong correlation between two variables often reflects the influence of a third, unobserved variable called a lurking variable. In this case, national wealth likely affects both chocolate consumption (wealthier countries can afford luxury foods) and Nobel laureates (wealthier countries typically have better educational systems, research funding, and universities). This creates an indirect relationship that appears as correlation without any direct causal link. Option A commits the classic correlation-causation fallacy. Just because two variables move together doesn't mean one causes the other – eating chocolate doesn't make individuals smarter or more likely to win Nobel prizes. Option B confuses correlation strength with statistical significance. The value $r = 0.95$ tells us about the strength of the linear relationship, not about statistical significance, which requires additional information like sample size and hypothesis testing. Option C attempts reverse causation but still assumes a direct causal relationship exists. While this reverses the direction, it's still incorrectly assuming that Nobel prizes directly influence chocolate consumption rather than both being influenced by a third factor. Remember this key strategy: when you see high correlations between seemingly unrelated variables, always consider lurking variables. Look for underlying factors that could logically influence both variables simultaneously – this approach will help you avoid correlation-causation traps throughout the statistics portion of your exam.
A café manager models the number of hot chocolates sold per day, $N$ , based on the average daily temperature, $T$ , in degrees Celsius. The regression equation is $N = -3.5T + 80$ . What is the interpretation of the value -3.5 in this context?
Answer: The number of hot chocolates sold is expected to decrease by 3.5 for each one-degree Celsius increase in temperature. The slope of the regression line, -3.5, represents the change in the dependent variable ( $N$ ) for a one-unit increase in the independent variable ( $T$ ). Therefore, for each 1°C increase in temperature, the number of hot chocolates sold is predicted to change by -3.5, which means it decreases by 3.5. - A is incorrect because it misinterprets the ratio of the change. - C is incorrect because the slope is a rate of change, not a minimum value. - D is an incorrect interpretation that confuses the slope with a specific prediction at T=0.
Given the context, find the Pearson's correlation coefficient, $r$ .
Answer: -0.85 The Pearson's correlation coefficient, $r$ , is the square root of the coefficient of determination, $r^2$ . The sign of $r$ must match the sign of the slope of the regression line. In the context of a used car's age and value, as age ( $x$ ) increases, value ( $y$ ) decreases. This implies a negative relationship, so $r$ must be negative. $r = -\sqrt{0.7225} = -0.85$ . - C is incorrect because it uses the positive root, which would imply that older cars are more valuable. - B is incorrect as it is the value of $r^2$ , not $r$ . - D is an incorrect calculation of the square root.
Which statement best describes the reliability of these predictions?
Answer: The prediction at $x=15$ is an interpolation and is considered more reliable than the prediction at $x=30$ . When you encounter regression problems involving predictions, the key distinction is between interpolation (predicting within the data range) and extrapolation (predicting outside the data range). Since your original data had x-values from 5 to 20, any prediction within this range is interpolation, while predictions outside this range are extrapolation. For the prediction at $x = 15$ : Since 15 falls between 5 and 20, this is interpolation. The model has actual data points around this value, making the prediction reliable because you're working within the established pattern. For the prediction at $x = 30$ : Since 30 is well beyond the upper limit of 20, this is extrapolation. You're assuming the linear relationship continues far beyond your data, which may not be valid. Real-world relationships often change or break down outside the observed range. Answer A incorrectly calls the $x = 30$ prediction an interpolation when it's actually extrapolation. Answer B wrongly claims both predictions are extrapolations—the $x = 15$ prediction is clearly within the data range. Answer C suggests extrapolation is more reliable for long-term trends, which contradicts statistical principles that favor predictions within the data range. Answer D correctly identifies that $x = 15$ is interpolation and is more reliable than the extrapolation at $x = 30$ . Study tip: Always check whether prediction values fall within or outside your original data range. Interpolation (within range) is generally reliable, while extrapolation (outside range) becomes increasingly unreliable the further you move from your data.
Based on these models, which statement is the most accurate comparison of the teams' performance?
Answer: Team B starts with a higher base value, but Team A becomes more effective as the number of calls increases. To compare the models, we find their intersection: $150c + 500 = 120c + 800$ , which gives $c = 10$ . At $c = 10$ , both teams have the same sales value of $2000. Team B has a higher y-intercept (800 vs 500), representing better performance with fewer calls, but Team A has a higher slope (150 vs 120), meaning they gain more value per additional call. Within the valid domain [10, 40], Team A performs better for $c > 10$ due to their higher rate of improvement. - A is incorrect because at $c = 10$ , they are equal. - C is incorrect because Team A outperforms Team B for most of the valid domain. - D discusses extrapolation beyond the model's validity.
What was the actual measured concentration of the chemical at this depth?
Answer: 4.7 ppm The residual is defined as $e = y_{actual} - y_{predicted}$ . We can rearrange this to find the actual value: $y_{actual} = y_{predicted} + e$ . 1. First, calculate the predicted concentration at $d=10$ metres: $C_{predicted} = -0.12(10) + 5.4 = -1.2 + 5.4 = 4.2$ ppm. 2. The residual is given as $e = +0.5$ ppm. 3. Calculate the actual concentration: $C_{actual} = 4.2 + 0.5 = 4.7$ ppm. - A is incorrect; it results from subtracting the residual from the predicted value and also making a calculation error. - B is incorrect as this is the predicted value, not the actual measured value. - D is incorrect as it results from adding the residual to the y-intercept.
The company uses the model to predict its profit for the year 2030. Which of the following statements is the most accurate?
Answer: The prediction for 2030 is likely to be unreliable because it is an extrapolation far beyond the data range. The model was built using data for the years 2011 to 2021 ( $t=1$ to $t=11$ ). Predicting for the year 2030 ( $t=20$ ) is an extrapolation because it falls outside the range of the data used to create the model. Extrapolations are often unreliable because the linear trend may not continue indefinitely. - A is incorrect; 2030 is outside the data range, so this is extrapolation, not interpolation. - C is incorrect; a strong trend within the data range does not guarantee the trend will continue. - D is incorrect; a negative intercept can be valid (e.g., initial loss) and does not inherently make the model flawed for predictions within its range.
Which of the following statements about the new regression line, $x$ on $y$ , is true?
Answer: The new regression line is mathematically distinct from the first line. The regression line of $y$ on $x$ is different from the regression line of $x$ on $y$ , unless the data lies perfectly on a line ( $r = \pm 1$ ). The first line minimizes vertical distances (residuals in $y$ ), while the second line minimizes horizontal distances (residuals in $x$ ). Therefore, they are distinct lines. - A is incorrect. The slope of the $x$ on $y$ line is not simply the reciprocal of the $y$ on $x$ line slope. The new slope is given by $r \frac{\sigma_x}{\sigma_y}$ , while the old slope was $r \frac{\sigma_y}{\sigma_x}$ . Their product is $r^2$ , not 1. - B is incorrect. The correlation coefficient $r$ is symmetric; it remains the same regardless of which variable is treated as independent. - D is incorrect. There's no simple relationship that guarantees the new intercept is -1.5. One cannot determine the new line without more information.
Calculate the residual for this student.
Answer: -1.0 The residual is the difference between the observed value and the predicted value (residual = observed $y$ - predicted $y$ ). 1. First, find the predicted score using the regression equation: $y_{predicted} = 0.85(80) + 12 = 68 + 12 = 80$ . 2. The observed score is $y_{observed} = 79$ . 3. Calculate the residual: $79 - 80 = -1.0$ . - B is incorrect due to a sign error (predicted - observed). - C is incorrect as it is the y-intercept of the model. - D is incorrect as it is the predicted value, not the residual.
Which of the following statements correctly describes the residual for this laptop?
Answer: The residual is $-\$30$ , meaning the model overestimated the laptop's price. First, calculate the predicted price: $P_{predicted} = 55(16) + 450 = 880 + 450 = 1330$ . The actual price is $P_{actual} = 1300$ . The residual is $P_{actual} - P_{predicted} = 1300 - 1330 = -30$ . A negative residual means the predicted value was higher than the actual value, so the model overestimated the price. - B is incorrect because the residual is negative, not positive, and it misinterprets the meaning of the sign. - C is incorrect because it gives the predicted price, not the residual. - D is incorrect because it is the value of the independent variable.
What is the mean of the y-values, $\bar{y}$ ?
Answer: 47.5 A fundamental property of the least squares regression line is that it always passes through the point corresponding to the means of the two variables, $(\bar{x}, \bar{y})$ . Therefore, we can find $\bar{y}$ by substituting $\bar{x}$ into the regression equation. $\bar{y} = -4.2(\bar{x}) + 110.5$ $\bar{y} = -4.2(15) + 110.5$ $\bar{y} = -63 + 110.5$ $\bar{y} = 47.5$ - A is incorrect; it assumes the means are equal. - C is incorrect; it is the result of the multiplication part of the calculation, ignoring the addition. - D is incorrect; it results from incorrectly adding 63 instead of subtracting.
Based on this information, which model provides the best fit for the data?
Answer: Model 3, because its correlation coefficient has the greatest magnitude. The best linear fit is indicated by the Pearson's correlation coefficient, $r$ , whose absolute value is closest to 1. We compare the magnitudes: $|0.82| = 0.82$ , $|0.91| = 0.91$ , $|-0.95| = 0.95$ , and $|0.88| = 0.88$ . The value with the greatest magnitude is -0.95, which corresponds to Model 3. This indicates the strongest linear relationship is between $\log(y)$ and $x$ . - A is incorrect because the original data may not have the most linear relationship. - B is incorrect because the sign of $r$ indicates direction, not strength of fit. A value of -0.95 indicates a stronger fit than 0.91. - D is incorrect; transforming both variables does not automatically make a model better.
The least squares regression line is the line that best fits a set of bivariate data. Which property is unique to this specific line?
Answer: It minimizes the sum of the squared vertical distances from the data points to the line. The defining characteristic of the least squares regression line is that it minimizes the sum of the squares of the residuals (the vertical distances between the observed y-values and the predicted y-values on the line). This is why it is called the 'least squares' line. - A is incorrect; the line may not pass through any of the data points. - C is incorrect; the sum of the residuals for a least squares regression line is always zero, not one. - D describes a different type of regression (Least Absolute Deviations regression), not the standard least squares method.
What is the correct interpretation of the $y$ -intercept of this equation?
Answer: The model predicts the plant was 8 cm tall at $t=0$ , but this is an unreliable extrapolation as it is outside the data range. The $y$ -intercept of a regression line is the predicted value of the dependent variable when the independent variable is zero. Here, it is the height at $t=0$ . However, the model was built using data from $t=5$ to $t=30$ . Predicting a value for $t=0$ is an extrapolation, meaning we are making a prediction outside the range of our data. Such predictions are often unreliable. - A is incorrect because it fails to recognize the unreliability of the extrapolation. - C is incorrect because the intercept is a predicted starting value, not an observed minimum. - D is incorrect because it misinterprets the intercept as the slope (rate of change).
How will the inclusion of this new data point, which is an outlier that does not fit the trend, likely affect the Pearson's correlation coefficient, $r$ ?
Answer: It will cause $r$ to move closer to 0. The Pearson's correlation coefficient, $r$ , measures the strength of a linear relationship. An outlier that does not follow the established trend will weaken the observed correlation. Since the original correlation was positive, a weaker positive correlation is represented by a value of $r$ closer to 0. - A is incorrect because this type of outlier weakens, not strengthens, the correlation. - C is incorrect because significant outliers can have a substantial effect on $r$ . - D is incorrect because while the outlier weakens the positive correlation, it is unlikely to be influential enough to make the overall correlation negative unless the original correlation was very weak.

Start this quiz