- \n
- \\(\\hat{Y}_i\\) is the predicted value of the dependent variable for observation \\(i\\)<\/li>\n
- \\(\\hat{b}_0\\) is the estimated intercept<\/li>\n
- \\(\\hat{b}_1\\) is the estimated slope coefficient<\/li>\n
- \\(X_i\\) is the observed value of the independent variable for observation \\(i\\)<\/li>\n<\/ul>\n
For forecasting, we usually write the predicted value as:<\/p>\n
$$\\hat{Y}_f = \\hat{b}_0 + \\hat{b}_1 X_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f\\) is the predicted or forecasted value of the dependent variable<\/li>\n
- \\(X_f\\) is the value of the independent variable used for the forecast<\/li>\n
- \\(\\hat{b}_0\\) is the estimated intercept<\/li>\n
- \\(\\hat{b}_1\\) is the estimated slope coefficient<\/li>\n<\/ul>\n
The predicted value is the point estimate. It is the model\u2019s best estimate of \\(Y\\) at a chosen value of \\(X\\), conditional on the fitted model. It is not a prediction interval and should not be interpreted as a range.<\/p>\n
Example 1: Calculating a Predicted Value<\/strong><\/p>\n
Assume we have a regression of monthly returns on a publicly listed battery-storage index against monthly returns on a battery-metals price index. The estimated regression equation is:<\/p>\n
$$\\hat{R}_{\\text{Storage}} = 0.011 + 0.42R_{\\text{Metals}}$$<\/p>\n
A commodities analyst expects the battery-metals index return to be -2.0% next month. The objective is to calculate the predicted return on the battery-storage index.<\/p>\n
Start by quoting the forecast equation:<\/p>\n
$$\\hat{Y}_f = \\hat{b}_0 + \\hat{b}_1 X_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f = \\hat{R}_{\\text{Storage}}\\), the predicted monthly return on the battery-storage index<\/li>\n
- \\(\\hat{b}_0 = 0.011\\), the estimated intercept<\/li>\n
- \\(\\hat{b}_1 = 0.42\\), the estimated slope coefficient<\/li>\n
- \\(X_f = R_{\\text{Metals}} = -0.020\\), the forecasted return on the battery-metals price index<\/li>\n<\/ul>\n
Substitute the values:<\/p>\n
$$ \\begin{align*}
\n\\hat{R}_{\\text{Storage}} &= 0.011 + 0.42(-0.020) \\\\
\n\\hat{R}_{\\text{Storage}} &= 0.011 – 0.0084 = 0.0026 = 0.26\\% \\end{align*} $$<\/p>\nThe predicted return on the battery-storage index is 0.26%<\/strong>. The positive intercept partly offsets the negative contribution from the metals return forecast.<\/p>\n
Standard Error of the Estimate<\/strong><\/p>\n
The standard error of the estimate, \\(s_e\\), measures the typical distance between observed values of the dependent variable and the values predicted by the regression model. It is also called the standard error of the regression or root mean squared error.<\/p>\n
For simple linear regression:<\/p>\n
$$s_e = \\sqrt{\\frac{SSE}{n-2}}$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(s_e\\) is the standard error of the estimate<\/li>\n
- \\(SSE\\) is the sum of squared errors<\/li>\n
- \\(n\\) is the number of observations<\/li>\n
- \\(n – 2\\) is the error degrees of freedom in simple linear regression<\/li>\n<\/ul>\n
The standard error of the estimate is in the same units as the dependent variable. If the dependent variable is monthly returns expressed in decimal form, \\(s_e\\) is also a monthly return in decimal form.<\/p>\n
A smaller \\(s_e\\) indicates that observed values tend to be closer to the fitted line. It does not guarantee that future predictions will be accurate, but it is an important input in the standard error of the forecast and prediction intervals.<\/p>\n
Example 2: Calculating the Standard Error of the Estimate<\/strong><\/p>\n
Let’s say a regression of monthly consumer-discretionary sector returns on monthly consumer-sentiment changes uses 50 observations. The sum of squared errors is 0.0800. The objective is to calculate the standard error of the estimate.<\/p>\n
Start with the formula:<\/p>\n
$$s_e = \\sqrt{\\frac{SSE}{n-2}}$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(s_e\\) is the standard error of the estimate<\/li>\n
- \\(SSE = 0.0800\\) is the sum of squared errors<\/li>\n
- \\(n = 50\\) is the number of observations<\/li>\n
- \\(n – 2 = 48\\) is the error degrees of freedom<\/li>\n<\/ul>\n
Substitute the values:<\/p>\n
$$s_e = \\sqrt{\\frac{0.0800}{48}} = \\sqrt{0.0016667} = 0.0408$$<\/p>\n
The standard error of the estimate is 0.0408<\/strong>, or 4.08%<\/strong> if returns are expressed in decimal form. This means observed monthly sector returns typically deviate from the fitted values by about 4.08 percentage points.<\/p>\n
Standard Error of the Forecast<\/h3>\n
The predicted value is only the center of the forecast. The standard error of the forecast measures the uncertainty around a forecast for a particular value of \\(X_f\\).<\/p>\n
The standard error of the forecast is:<\/p>\n
$$s_f = s_e \\sqrt{1 + \\frac{1}{n} + \\frac{(X_f – \\bar{X})^2}{\\sum_{i=1}^n (X_i – \\bar{X})^2}}$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(s_f\\) is the standard error of the forecast<\/li>\n
- \\(s_e\\) is the standard error of the estimate<\/li>\n
- \\(n\\) is the number of observations used to estimate the regression<\/li>\n
- \\(X_f\\) is the value of the independent variable used for the forecast<\/li>\n
- \\(\\bar{X}\\) is the sample mean of the independent variable<\/li>\n
- \\(\\sum_{i=1}^n (X_i – \\bar{X})^2\\) is the total squared variation of the independent variable around its mean<\/li>\n<\/ul>\n
The formula contains three important parts inside the square root.<\/p>\n
The first term, 1, reflects the uncertainty of the actual future observation around the regression line. The second term, \\(1\/n\\), reflects estimation uncertainty from using a finite sample. The third term increases when \\(X_f\\) is far from \\(\\bar{X}\\). Forecasts near the center of the historical data are more reliable than forecasts far away from the data center.<\/p>\n
Figure 1: Prediction Intervals Reflect Uncertainty Around the Forecasted Dependent Variable<\/strong><\/p>\n
![\"\"](\"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg\")
<\/p>\nFigure 1 shows that the prediction interval is narrowest near the mean of the independent variable and widens as the forecast value of X moves farther away from the center of the observed data. This is why extrapolation is dangerous. A forecast made far outside the historical range is exposed to greater uncertainty even before considering economic regime changes.<\/p>\n
Figure 2: Drivers of the Standard Error of the Forecast<\/strong><\/p>\n
![\"\"](\"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img2.jpg\")
<\/p>\nFigure 2 summarizes the drivers of forecast uncertainty. Better model fit lowers \\(s_e\\) and therefore lowers \\(s_f\\). A larger sample size reduces estimation uncertainty. A forecast made near \\(\\bar{X}\\) is more precise than a forecast far from \\(\\bar{X}\\). Greater variation in observed \\(X\\) values improves precision in estimating the slope.<\/p>\n
Example 3: Calculating the Standard Error of the Forecast<\/strong><\/p>\n
Imagine a regression of monthly infrastructure-equity returns on monthly inflation-surprise data. The model has the following information:<\/p>\n
$$
\n\\begin{array}{l|l}
\n\\textbf{Input} & \\textbf{Value} \\\\ \\hline
\n\\text{Standard error of the estimate, } s_e & 3.20\\% \\\\ \\hline
\n\\text{Number of observations, } n & 49 \\\\ \\hline
\n\\text{Mean of independent variable, } \\bar{X} & 1.50\\% \\\\ \\hline
\n\\text{Forecast value, } X_f & 4.50\\% \\\\ \\hline
\n\\sum_{i=1}^n (X_i – \\bar{X})^2 & 0.0380
\n\\end{array}
\n$$<\/p>\nThe objective is to calculate the standard error of the forecast.<\/p>\n
Start with the formula:<\/p>\n
$$s_f = s_e \\sqrt{1 + \\frac{1}{n} + \\frac{(X_f – \\bar{X})^2}{\\sum_{i=1}^n (X_i – \\bar{X})^2}}$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(s_f\\) is the standard error of the forecast<\/li>\n
- \\(s_e = 3.20\\% = 0.0320\\)<\/li>\n
- \\(n = 49\\)<\/li>\n
- \\(X_f = 4.50\\% = 0.0450\\)<\/li>\n
- \\(\\bar{X} = 1.50\\% = 0.0150\\)<\/li>\n
- \\(\\sum_{i=1}^n (X_i – \\bar{X})^2 = 0.0380\\)<\/li>\n<\/ul>\n
Calculate the distance term:<\/p>\n
$$X_f – \\bar{X} = 0.0450 – 0.0150 = 0.0300$$<\/p>\n
$$\\frac{(X_f – \\bar{X})^2}{\\sum_{i=1}^n (X_i – \\bar{X})^2} = \\frac{0.0300^2}{0.0380} = 0.0237$$<\/p>\n
Now calculate \\(s_f\\):<\/p>\n
$$ \\begin{align*} s_f & = 0.0320 \\sqrt{1 + \\frac{1}{49} + 0.0237} \\\\
\ns_f & = 0.0320 \\sqrt{1.0441} = 0.0327 \\end{align*} $$<\/p>\nThe standard error of the forecast is 3.27%<\/strong>. It is slightly larger than the standard error of the estimate because the forecast includes uncertainty around a future observation and because \\(X_f\\) is not exactly equal to \\(\\bar{X}\\).<\/p>\n
Prediction Intervals<\/h3>\n
A prediction interval gives a range of likely values for the dependent variable at a specified value of the independent variable. It is built around the predicted value:<\/p>\n
$$\\hat{Y}_f \\pm t_{\\alpha\/2} s_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f\\) is the predicted value of the dependent variable<\/li>\n
- \\(t_{\\alpha\/2}\\) is the two-tailed critical value from the t-distribution<\/li>\n
- \\(s_f\\) is the standard error of the forecast<\/li>\n
- \\(\\alpha\\) is the total probability outside the prediction interval<\/li>\n<\/ul>\n
The degrees of freedom are \\(n – 2\\) in simple linear regression because the model estimates two parameters: an intercept and a slope.<\/p>\n
A prediction interval is wider than a confidence interval for the conditional mean. A confidence interval for the mean estimates where the average value of \\(Y\\) lies for a given \\(X\\). A prediction interval estimates where an individual future observation may fall, so it includes more uncertainty.<\/p>\n
Example 4: Constructing a Prediction Interval<\/strong><\/p>\n
Using the previous infrastructure-equity example, assume the predicted monthly return is 1.20%, the standard error of the forecast is 3.27%, and the two-tailed 95% critical t-value is 2.01. The objective is to construct a 95% prediction interval.<\/p>\n
Start with the prediction interval formula:<\/p>\n
$$\\hat{Y}_f \\pm t_{\\alpha\/2} s_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f = 1.20\\%\\) is the predicted monthly return<\/li>\n
- \\(t_{\\alpha\/2} = 2.01\\) is the two-tailed 95% critical value<\/li>\n
- \\(s_f = 3.27\\%\\) is the standard error of the forecast<\/li>\n<\/ul>\n
Calculate the margin of error:<\/p>\n
$$2.01(3.27\\%) = 6.57\\%$$<\/p>\n
The prediction interval is:<\/p>\n
$$1.20\\% \\pm 6.57\\% = [-5.37\\%, 7.77\\%]$$<\/p>\n
\nThe 95% prediction interval is [-5.37%, 7.77%]<\/strong>. The wide interval shows that even a positive point forecast can be surrounded by substantial uncertainty.<\/p>\n
Why Forecasts Far from the Mean of X Are Less Precise<\/h3>\n
The forecast error formula contains the term:<\/p>\n
$$\\frac{(X_f – \\bar{X})^2}{\\sum_{i=1}^n (X_i – \\bar{X})^2}$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(X_f\\) is the forecast value of the independent variable<\/li>\n
- \\(\\bar{X}\\) is the sample mean of the independent variable<\/li>\n
- \\(\\sum_{i=1}^n (X_i – \\bar{X})^2\\) is the total squared variation in observed \\(X\\) values<\/li>\n<\/ul>\n<\/div>\n
The numerator grows when \\(X_f\\) moves farther from \\(\\bar{X}\\). The denominator grows when the sample has more spread in \\(X\\). A forecast at an extreme \\(X_f\\) is less precise because the estimated slope is being applied where the data provide less direct support.<\/p>\n
Example 5: Comparing Forecast Uncertainty Near and Far from \\(\\bar{X}\\)<\/strong><\/p>\n
Assume a regression has \\(s_e = 2.50\\%\\), \\(n = 64\\), and \\(\\sum (X_i – \\bar{X})^2 = 0.0400\\). Compare two forecast points:<\/p>\n
$$
\n\\begin{array}{l|c}
\n\\textbf{Forecast Point} & \\mathbf{X_f – \\bar{X}} \\\\
\n\\hline
\n\\text{Near the center} & 0.0050 \\\\
\n\\hline
\n\\text{Far from the center} & 0.0500
\n\\end{array}
\n$$<\/p>\nFor the near-center forecast:<\/p>\n
$$ \\begin{align*} s_f & = 0.0250 \\sqrt{1 + \\frac{1}{64} + \\frac{0.0050^2}{0.0400}} \\\\
\ns_f & = 0.0250 \\sqrt{1.0163} = 0.0252 = 2.52\\% \\end{align*} $$<\/p>\nFor the far-from-center forecast:<\/p>\n
$$ \\begin{align*} s_f & = 0.0250 \\sqrt{1 + \\frac{1}{64} + \\frac{0.0500^2}{0.0400}} \\\\
\ns_f &= 0.0250 \\sqrt{1.0781} = 0.0260 = 2.60\\% \\end{align*} $$<\/p>\nThe far-from-center forecast has the larger standard error. The difference may look small in this example, but it can become large when the forecast value is far outside the observed range or when the independent variable has little historical variation.<\/p>\n
Functional Forms in Simple Linear Regression<\/h3>\n
Not every financial or economic relationship is linear in the original variables. Some relationships are exponential, some show diminishing marginal effects, and others are best interpreted as elasticities. Ordinary least squares can still be useful if the variables are transformed so that the transformed relationship is linear.<\/p>\n
Common transformations include logarithms, reciprocals, squares, and differencing. This learning outcome focuses mainly on log-based functional forms.<\/p>\n
Figure 3: Functional Forms in Simple Linear Regression<\/strong><\/p>\n
![\"\"](\"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img3.jpg\")
<\/p>\nFigure 3 compares four common functional forms. The linear model uses \\(Y\\) and \\(X\\) directly. The log-lin model logs the dependent variable only. The lin-log model logs the independent variable only. The log-log model logs both variables and is especially useful when the slope is interpreted as an elasticity.<\/p>\n
$$ \\textbf{Table 1: Functional Forms and Coefficient Interpretation} \\\\
\n\\begin{array}{l|l|l|l}
\n{\\textbf{Functional} \\\\ \\textbf{Form}} & \\textbf{Regression Equation} & \\bf{\\text{Main Interpretation of } b_1} & \\textbf{Common Use} \\\\
\n\\hline
\n\\text{Linear} & Y_i = b_0 + b_1 X_i + \\epsilon_i & {\\text{Absolute change in Y} \\\\ \\text{for a one-unit change in X }} & \\text{Straight-line relationships} \\\\ \\hline
\n\\text{Log-lin} & \\ln Y_i = b_0 + b_1 X_i + \\epsilon_i & {\\text{Approximate percentage change in Y}\\\\ \\text{ for a one-unit change in X}} & \\text{Exponential growth or decay} \\\\ \\hline
\n\\text{Lin-log} & Y_i = b_0 + b_1 \\ln X_i + \\epsilon_i & {\\text{Absolute change in Y }\\\\ \\text{for a percentage change in X} }& \\text{Diminishing marginal effects} \\\\ \\hline
\n\\text{Log-log} & \\ln Y_i = b_0 + b_1 \\ln X_i + \\epsilon_i & {\\text{Elasticity: percentage change in Y} \\\\ \\text{for a 1% change in X} }& \\text{Proportional relationships} \\\\
\n\\end{array}
\n$$<\/p>\nTable 1 is essential for interpretation. The same numeric slope can mean very different things depending on which variable is logged.<\/p>\n
The Log-Lin Model<\/h3>\n
In a log-lin model, the dependent variable is transformed using the natural logarithm, while the independent variable remains in its original form:<\/p>\n
$$\\ln Y_i = b_0 + b_1 X_i + \\epsilon_i$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\ln Y_i\\) is the natural logarithm of the dependent variable<\/li>\n
- \\(X_i\\) is the independent variable<\/li>\n
- \\(b_0\\) is the intercept<\/li>\n
- \\(b_1\\) is the slope coefficient<\/li>\n
- \\(\\epsilon_i\\) is the error term<\/li>\n<\/ul>\n
The slope coefficient gives the approximate percentage change in \\(Y\\) for a one-unit increase in \\(X\\):<\/p>\n
$$\\%\\Delta Y \\approx 100b_1$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\%\\Delta Y\\) is the approximate percentage change in \\(Y\\)<\/li>\n
- \\(b_1\\) is the slope coefficient in the log-lin model<\/li>\n<\/ul>\n
The exact percentage change is:<\/p>\n
$$\\%\\Delta Y = 100\\left(e^{b_1} – 1\\right)$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(e\\) is the base of the natural logarithm<\/li>\n
- \\(b_1\\) is the slope coefficient<\/li>\n<\/ul>\n
Example 6: Interpreting a Log-Lin Model<\/strong><\/p>\n
Imagine a fixed-income analyst models the price of a zero-coupon bond as a function of time to maturity. The estimated model is:<\/p>\n
$$\\ln(P) = 4.75 – 0.035T$$<\/p>\n
The objective is to interpret the slope.<\/p>\n
Because the dependent variable is logged and the independent variable is not, this is a log-lin model. The approximate effect of a one-year increase in maturity is:<\/p>\n
$$100(-0.035) = -3.50\\%$$<\/p>\n
The exact effect is:<\/p>\n
$$100\\left(e^{-0.035} – 1\\right) = -3.44\\%$$<\/p>\n
The slope indicates that, holding the model structure constant, each additional year to maturity is associated with an approximate 3.50% decline in bond price. The exact interpretation is a 3.44% decline.<\/p>\n
The Lin-Log Model<\/h3>\n
In a lin-log model, the dependent variable remains in its original form, while the independent variable is transformed using the natural logarithm:<\/p>\n
$$Y_i = b_0 + b_1 \\ln X_i + \\epsilon_i$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(Y_i\\) is the dependent variable<\/li>\n
- \\(\\ln X_i\\) is the natural logarithm of the independent variable<\/li>\n
- \\(b_0\\) is the intercept<\/li>\n
- \\(b_1\\) is the slope coefficient<\/li>\n
- \\(\\epsilon_i\\) is the error term<\/li>\n<\/ul>\n
In a lin-log model, a 1% increase in \\(X\\) changes \\(Y\\) by approximately \\(0.01b_1\\) units.<\/p>\n
Example 7: Interpreting a Lin-Log Model<\/strong><\/p>\n
Assume an analyst studies how digital advertising expenditure affects sales growth. The estimated model is:<\/p>\n
$$\\widehat{\\text{Sales Growth}} = 0.030 + 0.080\\ln(\\text{Advertising Spend})$$<\/p>\n
Sales growth is measured in decimal form. The objective is to interpret the slope.<\/p>\n
For a 1% increase in advertising spend, the approximate change in sales growth is:<\/p>\n
$$0.01(0.080) = 0.0008$$<\/p>\n
In percentage-point terms, this is:<\/p>\n
$$0.0008 = 0.08 \\text{ percentage points}$$<\/p>\n
The model indicates that a 1% increase in advertising spend is associated with an increase of approximately 0.08 percentage points in predicted sales growth. This type of model is useful when additional spending has diminishing marginal effects.<\/p>\n
The Log-Log Model<\/h3>\n
In a log-log model, both the dependent and independent variables are transformed using natural logarithms:<\/p>\n
$$\\ln Y_i = b_0 + b_1 \\ln X_i + \\epsilon_i$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\ln Y_i\\) is the natural logarithm of the dependent variable<\/li>\n
- \\(\\ln X_i\\) is the natural logarithm of the independent variable<\/li>\n
- \\(b_0\\) is the intercept<\/li>\n
- \\(b_1\\) is the slope coefficient<\/li>\n
- \\(\\epsilon_i\\) is the error term<\/li>\n<\/ul>\n
The slope coefficient is an elasticity. It gives the approximate percentage change in \\(Y\\) for a 1% change in \\(X\\).<\/p>\n
Example 8: Interpreting a Log-Log Model<\/strong><\/p>\n
Consider a sustainability analyst modeling electricity demand as a function of real GDP. The estimated regression is:<\/p>\n
\n$$\\ln(\\text{Electricity Demand}) = 1.10 + 0.75\\ln(\\text{Real GDP})$$<\/p>\n
The objective is to interpret the slope.<\/p>\n
Because both variables are logged, this is a log-log model. The slope coefficient of 0.75 means that a 1% increase in real GDP is associated with a 0.75% increase in predicted electricity demand.<\/p>\n
The coefficient is not a percentage-point effect; it is an elasticity.<\/p>\n
Choosing an Appropriate Functional Form<\/strong><\/p>\n
The correct functional form should be guided by economic reasoning and diagnostic evidence. A higher \\(R^2\\) alone is not enough. The analyst should ask whether the transformed model makes financial sense and whether residual plots show fewer patterns.<\/p>\n
$$ \\textbf{Table 2: Choosing a Functional Form} \\\\
\n \\begin{array}{l|l|l}
\n \\textbf{Situation} & {\\textbf{Functional Form} \\\\ \\text{to Consider}} & \\textbf{Reason} \\\\
\n \\hline
\n \\text{Relationship appears straight-line} & \\text{Linear} & \\text{Slope is easy to interpret as an absolute change} \\\\ \\hline
\n {\\text{Dependent variable grows} \\\\ \\text{exponentially}} & \\text{Log-lin} & \\text{Logging } Y \\text{ can linearize exponential growth} \\\\ \\hline
\n {\\text{Independent variable has} \\\\ \\text{diminishing marginal impact}} & \\text{Lin-log} & \\text{Logging } X \\text{ can capture slowing effects} \\\\ \\hline
\n {\\text{Percentage changes in both} \\\\ \\text{variables matter}} & \\text{Log-log} & \\text{Slope can be interpreted as elasticity} \\\\ \\hline
\n \\text{Residuals show curvature} & {\\text{Transformation or} \\\\ \\text{nonlinear} \\\\ \\text{specification}} & \\text{A straight-line model may be misspecified}
\n \\end{array}
\n $$\n <\/p>\nA transformation can improve fit, but it changes interpretation. A forecast from a log model may also require converting back from log units to the original units. Candidates should be careful not to compare coefficients across models when the dependent variables are in different forms.<\/p>\n
Interpretation and Evaluation<\/h3>\n
Several practical principles follow:<\/p>\n
- \n
- The predicted value is a point estimate, not a range.<\/li>\n
- The standard error of the estimate measures typical fitted-value error in the units of the dependent variable.<\/li>\n
- The standard error of the forecast is larger when model fit is weaker, sample size is smaller, \\(X_f\\) is farther from \\(\\bar{X}\\), or variation in \\(X\\) is limited.<\/li>\n
- Prediction intervals are built around predicted values using the standard error of the forecast and a t-critical value.<\/li>\n
- Prediction intervals are wider than confidence intervals for the conditional mean because they describe an individual future observation.<\/li>\n
- Log-lin models are useful when \\(Y\\) changes by a percentage for a unit change in \\(X\\).<\/li>\n
- Lin-log models are useful when \\(Y\\) changes by an absolute amount for a percentage change in \\(X\\).<\/li>\n
- Log-log models are useful when the slope should be interpreted as elasticity.<\/li>\n
- Residual analysis and economic logic should guide the choice of functional form.<\/li>\n<\/ul>\n
Question<\/h2>\n
A regression of monthly fintech-index returns on monthly payment-volume growth produces the following estimated equation:<\/p>\n
$$\\hat{R}_{\\text{Fintech}} = 0.006 + 0.50G_{\\text{Payments}}$$<\/p>\n
\nFor next month, payment-volume growth is forecasted to be 3.0%. The standard error of the forecast is 4.00%, and the relevant two-tailed 95% t-critical value is 2.02.<\/p>\n
The 95% prediction interval for next month’s fintech-index return is closest to:<\/p>\n
- \n
- [-7.16%, 8.84%]<\/p>\n
[-5.98%, 10.18%]<\/p>\n
[-5.58%, 9.62%]<\/p>\n
Answer Explanation<\/strong><\/p>\n
The correct answer is B<\/strong>.<\/p>\n
First calculate the predicted value using the forecast equation:<\/p>\n
$$\\hat{Y}_f = \\hat{b}_0 + \\hat{b}_1 X_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f = \\hat{R}_{\\text{Fintech}}\\), the predicted fintech-index return<\/li>\n
- \\(\\hat{b}_0 = 0.006\\), the estimated intercept<\/li>\n
- \\(\\hat{b}_1 = 0.50\\), the estimated slope coefficient<\/li>\n
- \\(X_f = G_{\\text{Payments}} = 0.030\\), the forecasted payment-volume growth<\/li>\n<\/ul>\n
Calculate:<\/p>\n
$$\\hat{R}_{\\text{Fintech}} = 0.006 + 0.50(0.030) = 0.021 = 2.10\\%$$<\/p>\n
Now quote the prediction interval formula:<\/p>\n<\/div>\n
\n$$\\hat{Y}_f \\pm t_{\\alpha\/2} s_f$$<\/p>\n
Where:<\/strong><\/p>\n
- \n
- \\(\\hat{Y}_f = 2.10\\%\\)<\/li>\n
- \\(t_{\\alpha\/2} = 2.02\\)<\/li>\n
- \\(s_f = 4.00\\%\\)<\/li>\n<\/ul>\n
Calculate the margin of error:<\/p>\n
$$2.02(4.00\\%) = 8.08\\%$$<\/p>\n
The prediction interval is:<\/p>\n
$$2.10\\% \\pm 8.08\\% = [-5.98\\%, 10.18\\%]$$<\/p>\n
Using exact decimal form, the predicted return is:<\/p>\n
$$0.006 + 0.50(0.030) = 0.0210$$<\/p>\n
The margin is:<\/p>\n
$$2.02(0.0400) = 0.0808$$<\/p>\n
Therefore:<\/p>\n
$$0.0210 \\pm 0.0808 = [-0.0598, 0.1018] = [-5.98\\%, 10.18\\%]$$<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"
Why Prediction Requires More Than a Fitted Line Regression is often used because analysts want to predict something: next month\u2019s asset return, a company\u2019s sales growth, a credit spread, a fund\u2019s factor exposure, or a macroeconomic variable. The fitted line…<\/p>\n","protected":false},"author":15,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-61969","post","type-post","status-publish","format-standard","hentry","category-uncategorized","blog-post","no-post-thumbnail","animate"],"acf":[],"yoast_head":"\n
Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes\" \/>\n<meta property=\"og:description\" content=\"Why Prediction Requires More Than a Fitted Line Regression is often used because analysts want to predict something: next month\u2019s asset return, a company\u2019s sales growth, a credit spread, a fund\u2019s factor exposure, or a macroeconomic variable. The fitted line...\" \/>\n<meta property=\"og:url\" content=\"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/\" \/>\n<meta property=\"og:site_name\" content=\"AnalystPrep | CFA\u00ae Exam Study Notes\" \/>\n<meta property=\"article:published_time\" content=\"2026-07-23T12:58:05+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-07-23T12:58:07+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"737\" \/>\n\t<meta property=\"og:image:height\" content=\"385\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Kosikos Tuitoek\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Kosikos Tuitoek\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"15 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/\"},\"author\":{\"name\":\"Kosikos Tuitoek\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/#\\\/schema\\\/person\\\/73df713e3b6e82ee139e1eff20cebe20\"},\"headline\":\"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression\",\"datePublished\":\"2026-07-23T12:58:05+00:00\",\"dateModified\":\"2026-07-23T12:58:07+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/\"},\"wordCount\":3240,\"image\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/cfa_L10_c_img1.jpg\",\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/\",\"url\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/\",\"name\":\"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/cfa_L10_c_img1.jpg\",\"datePublished\":\"2026-07-23T12:58:05+00:00\",\"dateModified\":\"2026-07-23T12:58:07+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/#\\\/schema\\\/person\\\/73df713e3b6e82ee139e1eff20cebe20\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#primaryimage\",\"url\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/cfa_L10_c_img1.jpg\",\"contentUrl\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/cfa_L10_c_img1.jpg\",\"width\":737,\"height\":385},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/uncategorized\\\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/#website\",\"url\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/\",\"name\":\"AnalystPrep | CFA\u00ae Exam Study Notes\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/#\\\/schema\\\/person\\\/73df713e3b6e82ee139e1eff20cebe20\",\"name\":\"Kosikos Tuitoek\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g\",\"caption\":\"Kosikos Tuitoek\"},\"url\":\"https:\\\/\\\/analystprep.com\\\/cfa-level-1-exam\\\/author\\\/kosikos-tuitoek-enockanalystprep-com\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/","og_locale":"en_US","og_type":"article","og_title":"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes","og_description":"Why Prediction Requires More Than a Fitted Line Regression is often used because analysts want to predict something: next month\u2019s asset return, a company\u2019s sales growth, a credit spread, a fund\u2019s factor exposure, or a macroeconomic variable. The fitted line...","og_url":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/","og_site_name":"AnalystPrep | CFA\u00ae Exam Study Notes","article_published_time":"2026-07-23T12:58:05+00:00","article_modified_time":"2026-07-23T12:58:07+00:00","og_image":[{"width":737,"height":385,"url":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg","type":"image\/jpeg"}],"author":"Kosikos Tuitoek","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Kosikos Tuitoek","Est. reading time":"15 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#article","isPartOf":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/"},"author":{"name":"Kosikos Tuitoek","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/#\/schema\/person\/73df713e3b6e82ee139e1eff20cebe20"},"headline":"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression","datePublished":"2026-07-23T12:58:05+00:00","dateModified":"2026-07-23T12:58:07+00:00","mainEntityOfPage":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/"},"wordCount":3240,"image":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#primaryimage"},"thumbnailUrl":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg","inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/","url":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/","name":"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression - AnalystPrep | CFA\u00ae Exam Study Notes","isPartOf":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/#website"},"primaryImageOfPage":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#primaryimage"},"image":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#primaryimage"},"thumbnailUrl":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg","datePublished":"2026-07-23T12:58:05+00:00","dateModified":"2026-07-23T12:58:07+00:00","author":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/#\/schema\/person\/73df713e3b6e82ee139e1eff20cebe20"},"breadcrumb":{"@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#primaryimage","url":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg","contentUrl":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-content\/uploads\/2026\/07\/cfa_L10_c_img1.jpg","width":737,"height":385},{"@type":"BreadcrumbList","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/uncategorized\/predicted-values-standard-error-prediction-intervals-and-functional-forms-in-regression\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/analystprep.com\/cfa-level-1-exam\/"},{"@type":"ListItem","position":2,"name":"Predicted Values, Standard Error, Prediction Intervals, and Functional Forms in Regression"}]},{"@type":"WebSite","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/#website","url":"https:\/\/analystprep.com\/cfa-level-1-exam\/","name":"AnalystPrep | CFA\u00ae Exam Study Notes","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/analystprep.com\/cfa-level-1-exam\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/analystprep.com\/cfa-level-1-exam\/#\/schema\/person\/73df713e3b6e82ee139e1eff20cebe20","name":"Kosikos Tuitoek","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/8260edaa3f7ba04cf6b536b3f7fd769007ecb789b3289ac0cc4c3ab8b3f7f061?s=96&d=mm&r=g","caption":"Kosikos Tuitoek"},"url":"https:\/\/analystprep.com\/cfa-level-1-exam\/author\/kosikos-tuitoek-enockanalystprep-com\/"}]}},"_links":{"self":[{"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/posts\/61969","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/users\/15"}],"replies":[{"embeddable":true,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/comments?post=61969"}],"version-history":[{"count":19,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/posts\/61969\/revisions"}],"predecessor-version":[{"id":61991,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/posts\/61969\/revisions\/61991"}],"wp:attachment":[{"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/media?parent=61969"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/categories?post=61969"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/analystprep.com\/cfa-level-1-exam\/wp-json\/wp\/v2\/tags?post=61969"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
- [-7.16%, 8.84%]<\/p>\n