Binary Outcomes & Log-Odds - Biostatistics

A logistic regression model predicting treatment response includes gender (coefficient = 0.5, where 1 = female, 0 = male) and age (coefficient = 0.02 per year). If the intercept is -1.8, what is the odds ratio for treatment response comparing a 50-year-old female to a 40-year-old male?
Answer: $e^{(0.5 + 0.02 \times 10)} = e^{0.7} = 2.01$ When comparing two groups in logistic regression, you need to calculate the difference in their log-odds, then exponentiate to get the odds ratio. The key is identifying what differs between your comparison groups and accounting for all those differences simultaneously. For this comparison, you're contrasting a 50-year-old female versus a 40-year-old male. The logistic regression equation is: log-odds = -1.8 + 0.5(gender) + 0.02(age). The intercept cancels out when calculating odds ratios, so you only need the coefficient differences. The female is coded as 1 (vs. 0 for male), contributing +0.5 to her log-odds. She's also 10 years older (50 vs. 40), contributing +0.02 × 10 = +0.2 to her log-odds. The total difference in log-odds is 0.5 + 0.2 = 0.7, so the odds ratio is $e^{0.7} = 2.01$ . This makes A correct. B reaches the same numerical answer but uses an unnecessarily complicated approach by separating the gender and age effects before multiplying. While mathematically equivalent due to exponent rules, it's less efficient. C only accounts for the gender difference ( $e^{0.5} = 1.65$ ) but completely ignores the 10-year age difference between the two individuals. D incorrectly uses the 50-year-old's absolute age rather than the 10-year age difference between groups, leading to $e^{0.02 \times 50} = e^{1.0}$ instead of $e^{0.02 \times 10} = e^{0.2}$ . Study tip: For logistic regression comparisons, always calculate the difference in predictors between groups, not their absolute values. Focus on what changes, not what stays the same.
A researcher reports that increasing age by 10 years corresponds to a log-odds change of 0.693 for developing hypertension. If a 40-year-old has a 25% probability of hypertension, what is the approximate probability for a 50-year-old?
Answer: 40% For the 40-year-old, odds = 0.25/0.75 = 0.333. The log-odds change of 0.693 means OR = exp(0.693) = 2.0. For the 50-year-old, odds = 0.333 × 2.0 = 0.667. Converting to probability: 0.667/(1+0.667) = 0.40 or 40%. Choice B uses additive rather than multiplicative scaling. Choice C applies the coefficient incorrectly. Choice D represents a common error in odds-to-probability conversion.
Based on the logistic regression output above, which statement correctly describes the relationship between age and treatment response?
Answer: Each additional year of age decreases the odds of response by approximately 2.3% The coefficient -0.023 means OR = exp(-0.023) = 0.977. This represents a 1 - 0.977 = 0.023 or 2.3% decrease in odds per year. Choice B correctly calculates the odds ratio but misinterprets it as an increase. Choice C confuses odds with probability. Choice D appears to double the percentage, a common error when interpreting small coefficients.
In a logistic model, the log-odds coefficient for BMI is 0.04 per kg/m². A patient with BMI 25 has predicted log-odds of -1.5. What are the predicted log-odds for a patient with BMI 35?
Answer: $-1.5 + 0.04(35 - 25) = -1.1$ When you encounter logistic regression questions involving log-odds calculations, remember that coefficients represent the change in log-odds per unit change in the predictor variable. The key is understanding how to apply these coefficients to predict outcomes for new values. The correct approach uses the fundamental principle that log-odds change linearly with predictor variables. Since the BMI coefficient is 0.04 per kg/m², each unit increase in BMI increases the log-odds by 0.04. For a patient moving from BMI 25 to BMI 35 (a 10-unit increase), the log-odds increase by $0.04 \times 10 = 0.4$ . Adding this to the baseline log-odds: $-1.5 + 0.4 = -1.1$ . Answer A correctly applies this logic. Answer B incorrectly multiplies the coefficient by the new BMI value (35) rather than the change in BMI, ignoring that we already know the log-odds at BMI 25. Answer C makes a conceptual error by involving the exponential function and multiplication, confusing log-odds calculations with odds ratio transformations. Answer D commits the same mistake as B, multiplying by the absolute BMI value instead of the change, and coincidentally arrives at the same incorrect result through different arithmetic. Remember this pattern: in linear relationships like logistic regression's log-odds, when you know the outcome at one predictor value, calculate the new outcome by adding (coefficient × change in predictor). Don't multiply by absolute values or introduce exponential functions unless you're specifically converting between log-odds and odds.
A logistic regression model examining the relationship between smoking status and lung cancer yields a coefficient of 1.386 for the smoking variable. If the baseline probability of lung cancer for non-smokers is 0.02, what is the predicted probability of lung cancer for smokers?
Answer: 0.078 The coefficient 1.386 represents the log-odds ratio. First, calculate the baseline odds: 0.02/(1-0.02) = 0.0204. The odds ratio is exp(1.386) = 4.0. The new odds for smokers is 0.0204 × 4.0 = 0.0816. Converting back to probability: 0.0816/(1+0.0816) = 0.078. Choice B uses incorrect baseline odds calculation. Choice C applies the coefficient directly without proper transformation. Choice D confuses odds ratio with risk ratio.
A logistic regression model predicting treatment dropout has the equation: log-odds = -1.5 + 0.4(anxiety_score). At what anxiety score is the predicted probability of dropout equal to 80%?
Answer: Anxiety score = $\frac{\ln(0.8/0.2) + 1.5}{0.4} =$ 7.22 When you encounter logistic regression problems asking for a specific probability threshold, you need to work backwards from the probability to find the predictor value using the relationship between odds, log-odds, and probability. Starting with 80% probability, you first convert this to odds: odds = probability/(1-probability) = 0.8/0.2 = 4. Since the logistic regression equation gives you log-odds, you take the natural logarithm: log-odds = ln(4) ≈ 1.386. Now you can solve for the anxiety score using the given equation: log-odds = -1.5 + 0.4(anxiety_score). Substituting: 1.386 = -1.5 + 0.4(anxiety_score). Rearranging: anxiety_score = (1.386 + 1.5)/0.4 = 2.886/0.4 ≈ 7.22. This matches the setup in choice A: $\frac{\ln(0.8/0.2) + 1.5}{0.4}$ . Choice B incorrectly uses the probability (0.8) directly instead of converting to log-odds first—you can't substitute probability into a log-odds equation. Choice C takes the natural log of the probability rather than the odds, missing the critical step of converting 0.8 to odds of 4. Choice D ignores the probability requirement entirely and just solves for when the linear predictor equals zero. Study tip: Remember the sequence for logistic regression probability problems: probability → odds → log-odds → solve for predictor. The key conversion is odds = p/(1-p), then take the natural log. Always convert probabilities to log-odds when working with logistic regression equations.
In a logistic regression for disease diagnosis, a biomarker has a coefficient of 0.25 per unit. The baseline log-odds (biomarker = 0) is -4.0. At what biomarker level do the odds of disease equal 1.0?
Answer: Biomarker = 16 units When you encounter logistic regression problems involving odds, remember that you're working with the relationship between log-odds (logit) and the predictor variables. The logistic regression equation is: log-odds = intercept + coefficient × predictor. To find where odds equal 1.0, you need to determine when log-odds equals 0 (since ln(1) = 0). Setting up the equation: 0 = -4.0 + 0.25 × biomarker level. Solving for the biomarker level: 0.25 × biomarker = 4.0, so biomarker = 4.0 ÷ 0.25 = 16 units. Let's verify: At biomarker = 16, log-odds = -4.0 + 0.25(16) = -4.0 + 4.0 = 0. Since e⁰ = 1, the odds equal 1.0, confirming answer A is correct. Answer B (4 units) represents the value you get before dividing by the coefficient—this would give log-odds = -4.0 + 0.25(4) = -3.0, making odds = e⁻³ ≈ 0.05. Answer C (20 units) might result from incorrectly adding rather than dividing: -4.0 + 0.25(20) = 1.0, giving odds ≈ 2.7. Answer D (0.25 units) confuses the coefficient value with the biomarker level, yielding log-odds = -4.0 + 0.25(0.25) ≈ -3.94. Remember this key strategy: when odds equal 1.0, log-odds must equal 0. Set your logistic equation equal to zero and solve algebraically. This approach works for any logistic regression odds calculation.
A logistic model for predicting surgical success includes surgeon experience (coefficient = 0.05 per year) and patient age (coefficient = -0.03 per year). What is the odds ratio comparing a 30-year-old patient with a surgeon having 20 years experience to a 40-year-old patient with a surgeon having 10 years experience?
Answer: $e^{0.05 \times (20-10) + (-0.03) \times (30-40)} = e^{0.5 + 0.3} = e^{0.8} = 2.23$ When comparing odds between two groups in logistic regression, you need to calculate the difference in the linear predictor (log-odds) for each variable, then exponentiate to get the odds ratio. The logistic model is: log-odds = intercept + 0.05(surgeon experience) - 0.03(patient age). To compare two scenarios, you calculate the difference in their linear predictors. For the 30-year-old with 20-year surgeon versus the 40-year-old with 10-year surgeon, the difference is: Δ(log-odds) = [0.05 × (20-10)] + [(-0.03) × (30-40)] = 0.05 × 10 + (-0.03) × (-10) = 0.5 + 0.3 = 0.8 The odds ratio is $e^{0.8} = 2.23$ , making answer A correct. Answer B calculates the absolute log-odds for just one scenario ( $e^{0.05 × 20 + (-0.03) × 30}$ ) rather than the difference between scenarios. This gives you odds for one group, not a comparison between groups. Answer C only considers the surgeon experience difference ( $e^{0.05 × 10}$ ) while ignoring the patient age difference entirely. This incomplete calculation misses half the comparison. Answer D only accounts for the age difference using the wrong baseline ( $e^{(-0.03) × 10}$ ), ignoring surgeon experience and using an incorrect age comparison. Remember: odds ratios in logistic regression require calculating the difference in linear predictors for all relevant variables, then exponentiating. Always include every variable that differs between your comparison groups—partial calculations lead to wrong conclusions.
In a logistic model predicting hospital mortality, the coefficient for severity score is 0.15 per point. A patient with a severity score of 20 has a predicted probability of death of 0.6. What is the predicted probability for a patient with a severity score of 25?
Answer: $\frac{1}{1 + e^{-(\ln(0.6/0.4) + 0.15 \times 5)}} = 0.75$ When you encounter logistic regression problems, remember that these models predict probabilities using the logistic function, where changes in predictor variables affect the log-odds (logit) linearly, not the probability directly. The correct approach uses the logistic model's fundamental structure. Since the coefficient is 0.15 per point, increasing the severity score by 5 points (from 20 to 25) increases the log-odds by $0.15 \times 5 = 0.75$ . First, convert the known probability (0.6) to log-odds: $\ln(0.6/0.4) = \ln(1.5)$ . Then add the increase: $\ln(1.5) + 0.75$ . Finally, convert back to probability using the logistic function. Answer A correctly implements this: $\frac{1}{1 + e^{-(\ln(0.6/0.4) + 0.15 \times 5)}} = 0.75$ . Answer B incorrectly adds the coefficient change directly to the probability ( $0.6 + 0.75 = 1.35$ ), ignoring that logistic regression affects log-odds, not probabilities linearly. This also produces an impossible probability > 1. Answer C multiplies the probability by the exponential of the coefficient change ( $0.6 \times e^{0.75}$ ), which has no basis in logistic regression theory and yields an impossible result > 1. Answer D calculates the odds ratio incorrectly by multiplying initial odds by the exponential term, but then presents this as a probability rather than converting back properly. Study tip: In logistic regression, coefficients affect log-odds linearly. Always convert probability → log-odds → add change → convert back to probability. Never add coefficients directly to probabilities.
A logistic regression analysis shows that smoking status has a coefficient of 0.693. The baseline probability of lung disease for non-smokers is 0.1. What is the probability of lung disease for smokers?
Answer: $\frac{1}{1 + e^{-(\ln(0.1/0.9) + 0.693)}} = 0.182$ When you encounter logistic regression problems, you're working with the logit function that models the log-odds of an outcome. The key insight is that logistic regression coefficients represent changes in log-odds, and you need to convert between probabilities and odds to solve these problems properly. To find the probability for smokers, you first convert the baseline probability to odds, then apply the coefficient. For non-smokers with probability 0.1, the odds are $\frac{0.1}{1-0.1} = \frac{0.1}{0.9}$ . The log-odds (logit) is $\ln(0.1/0.9)$ . Since the smoking coefficient is 0.693, the log-odds for smokers becomes $\ln(0.1/0.9) + 0.693$ . Converting back to probability using the inverse logit function gives $\frac{1}{1 + e^{-(\ln(0.1/0.9) + 0.693)}} = 0.182$ , making A correct. B incorrectly treats the coefficient as an additive probability change, but logistic regression works on the log-odds scale, not the probability scale directly. C multiplies the baseline probability by $e^{0.693} = 2$ , which conceptually relates to the odds ratio but ignores that this multiplication should happen in odds space, not probability space. D appears to use some form of odds calculation but applies an incorrect formula that doesn't follow logistic regression principles. Remember: logistic regression coefficients affect log-odds, so always convert probabilities to odds, apply the coefficient, then convert back to probability using the logit function.
In a logistic regression, the log-odds of disease occurrence is modeled as -4.0 + 0.1(age). At what age does the predicted probability of disease equal 50%?
Answer: Age = 40 years This question tests your understanding of logistic regression and the relationship between log-odds and probability. When you see a logistic regression equation, remember that it models log-odds, not probability directly, and you'll need to convert between these measures. The logistic regression equation gives us: log-odds = -4.0 + 0.1(age). A 50% probability corresponds to log-odds of 0, because when P = 0.5, the odds are 1:1, and ln(1) = 0. Setting the equation equal to 0: 0 = -4.0 + 0.1(age). Solving for age: 0.1(age) = 4.0, so age = 40 years. Let's verify why each answer choice works or fails. Choice A (40 years) gives us log-odds = -4.0 + 0.1(40) = 0, which converts to exactly 50% probability using the formula $P = \frac{e^{0}}{1 + e^{0}} = \frac{1}{2} = 0.5$ . Choice B (45 years) produces log-odds = 0.5, giving a probability greater than 50%. Choice C (50 years) yields log-odds = 1.0, resulting in about 73% probability. Choice D (35 years) gives log-odds = -0.5, producing a probability less than 50%. Remember this key relationship: in logistic regression, 50% probability always occurs when log-odds equals zero. This creates a simple algebraic equation to solve. Also, the coefficient tells you the direction—positive coefficients mean probability increases with the predictor variable, as we see here with age.
A researcher models the log-odds of disease as: $\ln\left(\frac{p}{1-p}\right) = -3.2 + 0.4X_1 + 0.6X_2$ . If $X_1 = 2$ and $X_2 = 1$ , and $X_2$ increases to 3 while $X_1$ remains constant, by what factor do the odds change?
Answer: 3.32 When you encounter logistic regression problems involving changes in predictor variables, focus on how coefficients translate to odds ratios. In logistic regression, the coefficient for a variable represents the change in log-odds per unit increase in that variable. Let's calculate the odds at both conditions. Initially with $X_1 = 2$ and $X_2 = 1$ : $\ln\left(\frac{p}{1-p}\right) = -3.2 + 0.4(2) + 0.6(1) = -3.2 + 0.8 + 0.6 = -1.8$ So the initial odds are $e^{-1.8} = 0.165$ . After $X_2$ increases to 3 while $X_1$ stays at 2: $\ln\left(\frac{p}{1-p}\right) = -3.2 + 0.4(2) + 0.6(3) = -3.2 + 0.8 + 1.8 = -0.6$ The new odds are $e^{-0.6} = 0.549$ . The factor by which odds changed is $\frac{0.549}{0.165} = 3.32$ , confirming answer (B). Alternatively, since $X_2$ increased by 2 units and has coefficient 0.6, the odds multiply by $e^{0.6 \times 2} = e^{1.2} = 3.32$ . (A) 1.82 likely results from using $e^{0.6}$ , the odds ratio for just a 1-unit increase in $X_2$ . (C) 6.05 might come from incorrectly adding rather than multiplying odds ratios. (D) 2.48 could result from calculation errors or misapplying the exponential function. Strategy tip: Remember that in logistic regression, when a variable increases by $k$ units, the odds multiply by $e^{\beta \times k}$ where $\beta$ is the coefficient. This shortcut saves time compared to calculating odds at each condition separately.
A logistic regression model for predicting diabetes includes family history (coefficient = 1.1). The model intercept is -2.3. What is the difference in predicted probabilities between individuals with and without a family history of diabetes?
Answer: $\frac{1}{1+e^{-(-1.2)}} - \frac{1}{1+e^{-(-2.3)}} = 0.768 - 0.091 = 0.677$ When you encounter logistic regression problems asking for probability differences, you need to calculate the actual predicted probabilities using the logistic function, then find their difference. In logistic regression, predicted probability equals $\frac{1}{1+e^{-(\text{intercept} + \text{coefficient} \times \text{variable})}}$ . For someone WITH family history, the linear predictor is $-2.3 + 1.1(1) = -1.2$ , giving probability $\frac{1}{1+e^{-(-1.2)}} = 0.768$ . For someone WITHOUT family history, the linear predictor is $-2.3 + 1.1(0) = -2.3$ , giving probability $\frac{1}{1+e^{-(-2.3)}} = 0.091$ . The difference is $0.768 - 0.091 = 0.677$ , making A correct. Option B calculates the odds ratio minus 1, which gives you the multiplicative change in odds, not the probability difference. Option C appears to calculate some form of probability but uses an incorrect formula that doesn't account for the intercept—this would only be valid if the intercept were zero. Option D simply states the coefficient value, which represents the change in log-odds, not probabilities. The key trap here is confusing different measures: coefficients represent log-odds changes, exponentiating coefficients gives odds ratios, but probability differences require computing actual probabilities via the logistic function. Always remember that in logistic regression, you must transform through the logistic function to get interpretable probabilities, and include all relevant terms (intercept plus coefficients) in your linear predictor.
In a logistic regression predicting hospital readmission, the log-odds for patients with diabetes is -1.2 and for patients without diabetes is -2.5. What is the correct interpretation of the diabetes effect?
Answer: Diabetes increases the odds of readmission by a factor of 3.67 The difference in log-odds is -1.2 - (-2.5) = 1.3. The odds ratio is exp(1.3) = 3.67, meaning diabetes increases odds by a factor of 3.67. Choice B incorrectly calculates a percentage decrease. Choice C confuses odds ratio with probability difference. Choice D uses exp(1.0) instead of exp(1.3), representing a common calculation error.
A logistic regression model includes an interaction term between gender (male=1, female=0) and treatment (active=1, placebo=0). The coefficients are: Gender: 0.5, Treatment: 1.2, Gender×Treatment: -0.8. What is the log-odds of the outcome for males receiving active treatment, relative to females receiving placebo?
Answer: 0.9 When you encounter logistic regression with interaction terms, you need to carefully construct the linear predictor by including all relevant main effects and interactions for your specific covariate pattern. For males receiving active treatment, you have Gender=1 and Treatment=1. The linear predictor becomes: $\text{log-odds} = \beta_0 + \beta_{\text{Gender}} \times 1 + \beta_{\text{Treatment}} \times 1 + \beta_{\text{Gender×Treatment}} \times (1 \times 1)$ Since this asks for log-odds relative to females receiving placebo (the reference group where Gender=0 and Treatment=0), the intercept $\beta_0$ cancels out. You're left with: $\text{Relative log-odds} = 0.5(1) + 1.2(1) + (-0.8)(1) = 0.5 + 1.2 - 0.8 = 0.9$ This confirms answer D is correct. A) 2.5 likely comes from adding all coefficients without considering the negative interaction: 0.5 + 1.2 + 0.8 = 2.5. B) 1.7 results from ignoring the interaction term entirely: 0.5 + 1.2 = 1.7. C) 0.4 might stem from incorrectly calculating the interaction effect or mishandling the reference group comparison. The key insight is that interaction terms modify the effect of main predictors. Here, the negative interaction (-0.8) means the treatment effect is actually weaker in males than females. Always include interaction terms when both constituent variables are "turned on" (equal to 1), and remember that relative comparisons eliminate the intercept term.
In a study of surgical complications, the log-odds equation is: log-odds = -3.5 + 0.6(surgery_type) + 0.05(age), where surgery_type is coded as 0 for laparoscopic and 1 for open surgery. What is the odds ratio comparing open to laparoscopic surgery for patients of the same age?
Answer: $e^{0.6} = 1.82$ When you encounter a logistic regression equation asking for an odds ratio between two groups, you need to focus on how the coefficient translates to the odds ratio for that variable. In this log-odds equation, the coefficient for surgery_type is 0.6. Since surgery_type is coded as 0 for laparoscopic and 1 for open surgery, this coefficient represents the change in log-odds when moving from laparoscopic (0) to open surgery (1). To convert from log-odds to an odds ratio, you exponentiate the coefficient. Therefore, the odds ratio comparing open to laparoscopic surgery is $e^{0.6} = 1.82$ , making answer A correct. Let's examine why the other options are wrong. Answer B ( $e^{0.65} = 1.92$ ) incorrectly adds both coefficients together, but you only need the surgery_type coefficient since you're comparing surgery types at the same age. Answer C ( $e^{-2.9} = 0.055$ ) mistakenly uses the intercept plus the surgery coefficient, but the intercept represents baseline log-odds, not the comparison between groups. Answer D simply divides the coefficients ( $\frac{0.6}{0.05} = 12$ ), but this arithmetic operation has no meaning in logistic regression - you must exponentiate coefficients to get odds ratios. Remember this key principle: in logistic regression, the odds ratio for any variable equals $e^{\text{coefficient}}$ . The intercept and other variables don't factor into calculating a specific odds ratio when comparing groups, assuming other variables are held constant.
A logistic model predicting surgical complications includes BMI as a continuous variable with coefficient 0.08. If a patient with BMI 25 has a predicted probability of complications of 0.15, what is the predicted probability for a patient with BMI 30?
Answer: 0.22 When you encounter logistic regression problems, remember that coefficients represent the change in log-odds per unit increase in the predictor variable. The key is understanding how to translate between probabilities and odds. Starting with the given information: a patient with BMI 25 has a predicted probability of 0.15. First, convert this to odds: $\text{odds} = \frac{0.15}{1-0.15} = \frac{0.15}{0.85} = 0.176$ The log-odds for BMI 25 is: $\ln(0.176) = -1.737$ Since the BMI coefficient is 0.08, each unit increase in BMI increases log-odds by 0.08. For a BMI increase from 25 to 30 (5 units), the log-odds increases by: $5 \times 0.08 = 0.40$ The new log-odds is: $-1.737 + 0.40 = -1.337$ Converting back to odds: $e^{-1.337} = 0.263$ Finally, converting to probability: $P = \frac{0.263}{1 + 0.263} = 0.208 \approx 0.22$ Answer C (0.22) is correct. Answer A (0.27) might result from incorrectly adding the coefficient change directly to the probability. Answer B (0.18) could come from miscalculating the direction of change or making arithmetic errors. Answer D (0.20) is close but represents rounding errors in the conversion process. Remember: logistic regression works on the log-odds scale, so you must convert probabilities to log-odds, apply the coefficient, then convert back to probability. Never apply coefficients directly to probabilities.
A logistic model for predicting readmission has coefficients: intercept = -2.8, length_of_stay = 0.12 per day, comorbidities = 0.4 per condition. A patient with 5 days length of stay and 2 comorbidities has what probability of readmission?
Answer: $\frac{1}{1 + e^{-(-2.8 + 0.12 \times 5 + 0.4 \times 2)}} = \frac{1}{1 + e^{1.6}} = 0.202$ When you encounter logistic regression problems, remember that you're calculating the probability of an event using the logistic function: $P = \frac{1}{1 + e^{-(\text{linear combination})}}$ , where the linear combination includes all terms from your model. To find this patient's readmission probability, you need to substitute the given values into the complete logistic model. The linear combination is: intercept + (length_of_stay coefficient × days) + (comorbidities coefficient × conditions) = -2.8 + (0.12 × 5) + (0.4 × 2) = -2.8 + 0.6 + 0.8 = -1.6. Therefore: $P = \frac{1}{1 + e^{-(-1.6)}} = \frac{1}{1 + e^{1.6}} = 0.202$ Choice A correctly applies the full logistic model with all components and arrives at the right probability of 0.202. Choice B omits the intercept term (-2.8), a critical error since the intercept represents the baseline log-odds when all predictors equal zero. This gives an incorrect linear combination of 1.4 instead of -1.6. Choice C uses only the exponential term $e^{-1.6}$ without the proper logistic transformation. This calculates neither probability nor odds correctly. Choice D appears to use an alternative form of the logistic function but incorrectly applies it, yielding 0.168 instead of the correct 0.202. Study tip: Always double-check that you've included every term in your logistic model, especially the intercept. The intercept is often forgotten but represents crucial baseline information that significantly affects your probability calculations.
A logistic regression model yields the equation: log-odds = -2.5 + 0.8X, where X represents treatment status (0 = control, 1 = treatment). What is the probability of the outcome for a patient in the treatment group?
Answer: $\frac{1}{1 + e^{-(-1.7)}} = 0.845$ When you encounter a logistic regression equation, you're working with a model that predicts the probability of a binary outcome. The key is understanding how to transform the log-odds equation into an actual probability using the logistic function: $P = \frac{1}{1 + e^{-(\text{log-odds})}}$ . Given the equation log-odds = -2.5 + 0.8X, you need to substitute X = 1 for the treatment group. This gives you: log-odds = -2.5 + 0.8(1) = -1.7. Now apply the logistic transformation: $P = \frac{1}{1 + e^{-(-1.7)}} = \frac{1}{1 + e^{1.7}} = 0.845$ . This matches option A perfectly. Option B incorrectly uses only the intercept (-2.5), ignoring the treatment effect entirely. This would give you the probability for the control group, not the treatment group. Option C uses only the coefficient (0.8) without the intercept, which misses the baseline log-odds. Option D uses the correct log-odds calculation (-1.7) but applies the wrong probability formula. The expression $\frac{e^{-1.7}}{1 + e^{-1.7}}$ actually calculates the probability of the opposite outcome (failure rather than success). Remember this two-step process: first substitute all variable values into the log-odds equation, then transform using the standard logistic function. The most common error is forgetting to include all terms in the equation or mixing up the probability formula.
A logistic model predicts treatment success with log-odds = -1.2 + 0.8(dose). If the current dose gives a 40% success rate, what dose would be needed to achieve a 70% success rate?
Answer: Current dose + $\frac{\ln(70/30) - \ln(40/60)}{0.8} = \text{Current dose} + 1.46$ When you encounter logistic regression problems, remember that the relationship between probability and dose involves log-odds, not direct proportions. The logistic model gives you log-odds = -1.2 + 0.8(dose), where log-odds equals $\ln\left(\frac{p}{1-p}\right)$ . To solve this, you need to find the change in dose required to go from 40% to 70% success. First, convert probabilities to odds ratios: 40% success means odds = 40/60, and 70% success means odds = 70/30. The log-odds difference is $\ln(70/30) - \ln(40/60)$ . Since the coefficient for dose is 0.8, the required dose change is $\frac{\ln(70/30) - \ln(40/60)}{0.8}$ , which equals approximately 1.46. Option A correctly applies this log-odds transformation and accounts for the coefficient, giving the right dose increase of 1.46 units. Option B incorrectly treats this as a linear relationship, simply dividing the probability difference (0.7 - 0.4) by the coefficient. This ignores the logarithmic nature of logistic regression. Option C attempts to solve for an absolute dose value rather than the change needed, and doesn't properly account for the current baseline. Option D uses a simple ratio multiplication, completely missing that logistic models involve logarithmic transformations, not direct proportional relationships. Study tip: In logistic regression problems, always convert probabilities to odds first, then work with log-odds. Never treat probability changes as linear when dealing with logistic models.

Start this quiz