Advanced Empirical Methods: Discrete Variable Models

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/89

flashcard set

Earn XP

Description and Tags

Comprehensive practice flashcards covering Discrete Variable Models: Ordered Categorical Models (Ordered Logit and Probit), Unordered Categorical Models (Multinomial Logit), Count Data Models (Poisson, Poisson PML, and Negative Binomial), and Response Heterogeneity (HOPIT with Anchoring Vignettes).

90 Terms

1
New cards

What is an ordered categorical variable?

A discrete variable whose response categories have a clear, natural ordering, but whose values are not continuous (generally coded as y=1,2,…,Jy = 1, 2, \dots, J).

2
New cards

What are three examples of ordered categorical variables mentioned in the lecture?

Self-assessed health, risk preferences/attitudes, firm sales growth (low to high), and student course evaluations.

3
New cards

In the working mothers practical example, what dataset and sample size were used?

The 1977 and 1989 US General Social Survey (GSS), with a sample size of N=2,293N = 2{,}293.

4
New cards

How is the dependent variable warm defined and coded in the working mothers example?

Based on agreement with the statement that a working mother can establish just as warm and secure a relationship with her child: 1=Strongly Disagree (SD)1 = \text{Strongly Disagree (SD)}, 2=Disagree (D)2 = \text{Disagree (D)}, 3=Agree (A)3 = \text{Agree (A)}, and 4=Strongly Agree (SA)4 = \text{Strongly Agree (SA)}.

5
New cards

What are the two key explanatory variables in the baseline ordered model for warm working mothers?

x1=yr89x_1 = \text{yr89} (a dummy variable with 1977 as the reference category) and x2=edx_2 = \text{ed} (education in years).

6
New cards

Why can standard linear regression models not be used for ordered categorical outcomes?

Because the outcome variable is discrete and ordinal; numerical category codes reflect ranking rather than continuous quantitative differences or equal intervals.

7
New cards

How is the underlying continuous process modeled in ordered regression models?

Through a continuous latent variable y∗=β1x1+β2x2+uy^* = \beta_1 x_1 + \beta_2 x_2 + u, which determines the observed discrete response yy.

8
New cards

What is the mathematical relationship mapping the latent variable y∗y^* to the observed category y=jy = j?

y=j if τj−1≤y∗<τjy = j \text{ if } \tau_{j-1} \le y^* < \tau_j for j=1,…,Jj = 1, \dots, J, where τ0=−∞\tau_0 = -\infty, τJ=+∞\tau_J = +\infty, and τ1≤τ2≤⋯≤τJ−1\tau_1 \le \tau_2 \le \dots \le \tau_{J-1} are unknown thresholds to be estimated.

9
New cards

What distributional assumption on the error term uu distinguishes ordered probit from ordered logit?

In ordered probit, uu follows a standard normal distribution (u∼N(0,1)u \sim N(0,1)), whereas in ordered logit, uu follows a standard logistic distribution.

10
New cards

What is the formula for P(y=j∣x1,x2)P(y = j \mid x_1, x_2) in the ordered probit model?

P(y=j∣x1,x2)=Φ(τj−β1x1−β2x2)−Φ(τj−1−β1x1−β2x2)P(y = j \mid x_1, x_2) = \Phi(\tau_j - \beta_1 x_1 - \beta_2 x_2) - \Phi(\tau_{j-1} - \beta_1 x_1 - \beta_2 x_2), where Φ(⋅)\Phi(\cdot) is the standard normal cumulative distribution function (CDF).

11
New cards

In an ordered logit model, why does P(y=1∣x1,x2)P(y = 1 \mid x_1, x_2) simplify to a single fraction?

Because τ0=−∞\tau_0 = -\infty, causing exp⁡(τ0−β1x1−β2x2)=0\exp(\tau_0 - \beta_1 x_1 - \beta_2 x_2) = 0, so the subtracted term equals zero: P(y=1∣x1,x2)=exp⁡(τ1−β1x1−β2x2)1+exp⁡(τ1−β1x1−β2x2)P(y = 1 \mid x_1, x_2) = \frac{\exp(\tau_1 - \beta_1 x_1 - \beta_2 x_2)}{1 + \exp(\tau_1 - \beta_1 x_1 - \beta_2 x_2)}.

12
New cards

In an ordered logit model with J=4J = 4, why does P(y=4∣x1,x2)P(y = 4 \mid x_1, x_2) simplify to 1−exp⁡(τ3−β1x1−β2x2)1+exp⁡(τ3−β1x1−β2x2)1 - \frac{\exp(\tau_3 - \beta_1 x_1 - \beta_2 x_2)}{1 + \exp(\tau_3 - \beta_1 x_1 - \beta_2 x_2)}?

Because τ4=+∞\tau_4 = +\infty, which makes the leading term exp⁡(τ4−β1x1−β2x2)1+exp⁡(τ4−β1x1−β2x2)=1\frac{\exp(\tau_4 - \beta_1 x_1 - \beta_2 x_2)}{1 + \exp(\tau_4 - \beta_1 x_1 - \beta_2 x_2)} = 1.

13
New cards

In the working mothers ordered logit regression, what were the estimated coefficients for yr89 and ed?

β^yr89=0.5146\hat{\beta}_{\text{yr89}} = 0.5146 (z=6.50z = 6.50) and β^ed=0.1024\hat{\beta}_{\text{ed}} = 0.1024 (z=8.22z = 8.22).

14
New cards

What were the three estimated cut-points (/cut1, /cut2, /cut3) in the baseline ordered logit model for warm?

τ1=−0.5120\tau_1 = -0.5120, τ2=1.2401\tau_2 = 1.2401, and τ3=3.0286\tau_3 = 3.0286.

15
New cards

What is the test statistic for testing a single hypothesis H0:βk=β∗H_0: \beta_k = \beta^* in ordered logit or probit models?

z=β^k−β∗σ^β^k∼N(0,1)z = \frac{\hat{\beta}_k - \beta^*}{\hat{\sigma}_{\hat{\beta}_k}} \sim N(0,1) asymptotically.

16
New cards
<p>What econometric estimation concept is depicted in this figure?</p>

What econometric estimation concept is depicted in this figure?

Maximum Likelihood estimation, illustrating that the parameter estimate θ^ML\hat{\theta}_{ML} maximizes the log-likelihood function log⁡L\log L.

17
New cards

How is the total log-likelihood function log⁡L\log L defined across nn observations in an ordered model?

log⁡L=∑i=1nlog⁡f(yi∣x1i,x2i)\log L = \sum_{i=1}^n \log f(y_i \mid x_{1i}, x_{2i}), where f(yi∣⋅)=P(y=j∣⋅)f(y_i \mid \cdot) = P(y = j \mid \cdot) for the observed response category jj.

18
New cards

What is the formula and asymptotic distribution of the Likelihood Ratio Test (LRT) statistic?

LRT=2(log⁡LUR−log⁡LR)∼χ2(q)\text{LRT} = 2(\log L_{UR} - \log L_R) \sim \chi^2(q), where qq is the number of parameter restrictions.

19
New cards

In Stata, what command performs a Wald test of multiple coefficient restrictions after ologit?

The test command (e.g., test male white age prst), which evaluates restrictions using an asymptotic χ2(q)\chi^2(q) distribution.

20
New cards

How is McFadden's Pseudo-R2R^2 defined in ordered logit and probit models?

Pseudo-R2=1−log⁡Llog⁡L0\text{Pseudo-}R^2 = 1 - \frac{\log L}{\log L_0}, where log⁡L\log L is the log-likelihood of the estimated model and log⁡L0\log L_0 is the log-likelihood of an intercept-only model.

21
New cards

If β2>0\beta_2 > 0 in an ordered model, what are the unambiguous signs of the marginal effects on P(y=1)P(y = 1) and P(y=J)P(y = J)?

The marginal effect on P(y=1)P(y = 1) is strictly negative, and the marginal effect on the highest category P(y=J)P(y = J) is strictly positive.

22
New cards

Why can the sign of the marginal effect on intermediate categories (j=2,…,J−1j = 2, \dots, J-1) not be determined by the sign of βk\beta_k alone?

Because intermediate probabilities depend on shifts between two adjacent cut-points, meaning their marginal effects can be positive or negative depending on the specific values of the covariates.

23
New cards

What is the difference between an average marginal effect and a conditional marginal effect?

An average marginal effect calculates the marginal effect for every individual observation and averages them across the sample (margins, dydx(*)), while a conditional marginal effect evaluates the effect at specified values of the covariates (margins, dydx(*) at(...)).

24
New cards

What characterizes an unordered categorical variable?

A discrete outcome whose categories have no natural ordering or hierarchy (e.g., transportation mode, choice of healthcare provider, or firm innovation strategy).

25
New cards

How many sets of slope coefficients are estimated in a Multinomial Logit (MNL) model with MM categories?

M−1M - 1 sets of coefficients, because one category must be set as the base/reference category whose parameters are normalized to zero.

26
New cards

In the Portuguese healthcare provider example, what were the three categories of the variable choice?

1=no specialist consultations in the past 3 months (base outcome)1 = \text{no specialist consultations in the past 3 months (base outcome)}, 2=consultation only with public specialist(s)2 = \text{consultation only with public specialist(s)}, and 3=consultation with private specialist(s)3 = \text{consultation with private specialist(s)}.

27
New cards

What were the sample size and explanatory variables in the healthcare choice MNL example?

The sample size was N=22,772N = 22{,}772, with explanatory variables female (dummy variable) and educ (education in years).

28
New cards

What is the formula for the probability of the base category (y=1y = 1) in a 3-category MNL model?

P(y=1∣x1,x2)=11+exp⁡(β0,2+β1,2x1+β2,2x2)+exp⁡(β0,3+β1,3x1+β2,3x2)P(y = 1 \mid x_1, x_2) = \frac{1}{1 + \exp(\beta_{0,2} + \beta_{1,2}x_1 + \beta_{2,2}x_2) + \exp(\beta_{0,3} + \beta_{1,3}x_1 + \beta_{2,3}x_2)}.

29
New cards

What is the formula for the probability of an alternative category mm (m>1m > 1) in an MNL model?

P(y=m∣x1,x2)=exp⁡(β0,m+β1,mx1+β2,mx2)1+∑k=2Mexp⁡(β0,k+β1,kx1+β2,kx2)P(y = m \mid x_1, x_2) = \frac{\exp(\beta_{0,m} + \beta_{1,m}x_1 + \beta_{2,m}x_2)}{1 + \sum_{k=2}^M \exp(\beta_{0,k} + \beta_{1,k}x_1 + \beta_{2,k}x_2)}.

30
New cards

What does the sign of an MNL coefficient β1,m\beta_{1,m} indicate?

It indicates the sign of the effect of x1x_1 on the probability of choosing category mm relative to the base category (P(y=m)P(y=1)\frac{P(y=m)}{P(y=1)}), not on the absolute probability P(y=m)P(y=m).

31
New cards

Does a positive coefficient β1,m>0\beta_{1,m} > 0 guarantee that an increase in x1x_1 increases the absolute probability P(y=m)P(y = m)?

No; P(y=m)P(y = m) could decrease if the probabilities of other alternative categories increase by an even greater amount.

32
New cards

In the healthcare choice MNL regression, what were the estimated coefficients of female for public and private care?

For public care (m=2m = 2), β^female,2=0.1274\hat{\beta}_{\text{female}, 2} = 0.1274 (p=0.001p = 0.001); for private care (m=3m = 3), β^female,3=0.1416\hat{\beta}_{\text{female}, 3} = 0.1416 (p=0.026p = 0.026), both relative to no care.

33
New cards

In the healthcare choice MNL regression, what were the estimated coefficients of educ for public and private care?

For public care (m=2m = 2), β^educ,2=−0.0426\hat{\beta}_{\text{educ}, 2} = -0.0426 (p=0.000p = 0.000); for private care (m=3m = 3), β^educ,3=0.0608\hat{\beta}_{\text{educ}, 3} = 0.0608 (p=0.000p = 0.000), both relative to no care.

34
New cards

Why is it incorrect to state that β^female,2=0.1274\hat{\beta}_{\text{female}, 2} = 0.1274 means being female increases the probability of public care by 12.7 percentage points?

Because MNL regression coefficients reflect log-relative-risk changes relative to the base category, not direct percentage point changes in absolute probabilities.

35
New cards

How is the overall hypothesis that female has no effect on any healthcare choice tested in Stata?

By testing the joint hypothesis H0:βfemale,2=0 and βfemale,3=0H_0: \beta_{\text{female}, 2} = 0 \text{ and } \beta_{\text{female}, 3} = 0 using the command test [2]female [3]female.

36
New cards

What was the result of the joint Wald test for female in the healthcare MNL model?

χ2(2)=14.21\chi^2(2) = 14.21 with p=0.0008p = 0.0008, rejecting the null hypothesis of no gender effect at the 1%1\% significance level.

37
New cards
<p>In this Stata plot of predicted probabilities from an MNL model, how do the choices change as education increases?</p>

In this Stata plot of predicted probabilities from an MNL model, how do the choices change as education increases?

P(choice=1)P(\text{choice}=1) (no care) stays dominant around 0.80.8, P(choice=2)P(\text{choice}=2) (public care) decreases from around 0.190.19 to 0.070.07, and P(choice=3)P(\text{choice}=3) (private care) increases from around 0.030.03 to 0.150.15.

38
New cards

In the healthcare MNL model, what was the average marginal effect of female on P(choice=1) (no care)?

−0.0207-0.0207 (p<0.001p < 0.001), meaning being female decreases the probability of having no specialist consultation by approximately 2.072.07 percentage points on average.

39
New cards

In the healthcare MNL model, what was the average marginal effect of female on P(choice=2) (public care)?

+0.0149+0.0149 (p=0.002p = 0.002), meaning being female increases the probability of public specialist consultation by approximately 1.491.49 percentage points on average.

40
New cards

In the healthcare MNL model, what was the average marginal effect of female on P(choice=3) (private care)?

+0.0058+0.0058 (p=0.050p = 0.050), meaning being female increases the probability of private specialist consultation by approximately 0.580.58 percentage points on average.

41
New cards

What is the mathematical sum of the marginal effects of a covariate across all categories in an MNL model, and why?

The sum equals zero (∑m=1M∂P(y=m)∂xk=0\sum_{m=1}^M \frac{\partial P(y=m)}{\partial x_k} = 0), because the probabilities of all mutually exclusive categories must always sum to 1.

42
New cards

What Stata command calculates the average marginal effect of covariates on the third category in an MNL model?

margins, dydx(*) predict(outcome(3)).

43
New cards

In the healthcare MNL model, what were the average marginal effects of education on public vs private specialist visits?

Education had a negative marginal effect on public care (−0.0058-0.0058, p<0.001p < 0.001) and a positive marginal effect on private care (+0.0032+0.0032, p<0.001p < 0.001).

44
New cards

By default, how does Stata's mlogit command select the base category?

Stata automatically chooses the category with the highest frequency as the base outcome, unless specified otherwise by the user.

45
New cards

Can the sign of the marginal effect of xkx_k on P(y=m)P(y=m) differ from the sign of the coefficient βk,m\beta_{k,m} in an MNL model?

Yes, because the marginal effect depends on the coefficients and probabilities of all other alternatives, not just category mm.

46
New cards

What is a count variable?

A discrete, non-negative integer variable (y∈{0,1,2,… }y \in \{0, 1, 2, \dots\}) representing the number of times an event occurs.

47
New cards

What are four empirical examples of count variables presented in the lecture?

Number of doctor visits/diagnostic tests, alcoholic drinks per week, cigarettes smoked per day, and unpaid parking violations.

48
New cards

What was the empirical question studied by Fisman and Miguel (2005/2007) in their UN parking tickets paper?

Whether cultural norms of corruption affect compliance with laws in the absence of legal enforcement, using unpaid parking violations of UN diplomats in Manhattan who had diplomatic immunity until 2002.

49
New cards

In the UN parking violations sample (N=149N = 149), what were the sample mean and variance of violations?

The sample mean was 11.2811.28 and the sample variance was 337.93337.93.

50
New cards

What is the defining property of the Poisson distribution regarding its mean and variance?

Equidispersion: the conditional variance equals the conditional mean (E(y)=Var⁡(y)=μE(y) = \operatorname{Var}(y) = \mu).

51
New cards

How is the conditional expectation function E(y∣x)E(y \mid x) specified in Poisson and Negative Binomial models?

E(y∣x1,x2)=μ=exp⁡(β0+β1x1+β2x2)E(y \mid x_1, x_2) = \mu = \exp(\beta_0 + \beta_1 x_1 + \beta_2 x_2), ensuring that expected counts are strictly non-negative.

52
New cards

What is overdispersion in count data?

A condition where the conditional variance exceeds the conditional mean (Var⁡(y)>E(y)\operatorname{Var}(y) > E(y)).

53
New cards

What are the statistical consequences of overdispersion when estimating a count model using standard Poisson Maximum Likelihood?

The parameter estimates of β\beta and E(y∣x)E(y \mid x) remain consistent, but standard errors are biased downwards, producing inflated zz-statistics and spuriously small pp-values.

54
New cards

How does the Negative Binomial (NB) model specify the conditional variance to accommodate overdispersion?

Var⁡(y)=μ(1+αμ)=μ+αμ2\operatorname{Var}(y) = \mu(1 + \alpha \mu) = \mu + \alpha \mu^2, where α>0\alpha > 0 is the overdispersion parameter.

55
New cards

Under what condition does the Negative Binomial model simplify to the standard Poisson model?

When the overdispersion parameter α=0\alpha = 0, which makes Var⁡(y)=μ(1+0)=μ=E(y)\operatorname{Var}(y) = \mu(1 + 0) = \mu = E(y).

56
New cards

What is Poisson Pseudo-Maximum Likelihood (Poisson PML)?

Estimating a Poisson regression model using robust (sandwich) standard errors (robust option in Stata) to provide valid standard errors without requiring the equidispersion assumption E(y)=Var⁡(y)E(y) = \operatorname{Var}(y).

57
New cards

How do Poisson PML and Negative Binomial regression compare in estimating probabilities P(y=j)P(y = j)?

Negative Binomial specifies a full distribution and is good at estimating individual probabilities P(y=j)P(y = j), whereas Poisson PML only correctly models the conditional mean and cannot reliably estimate P(y=j)P(y = j) under overdispersion.

58
New cards

In the Poisson ML estimation of the parking tickets data, what were the coefficients and z-statistics for staff and corruption?

For staff, β^=0.0164\hat{\beta} = 0.0164 (z=9.57z = 9.57); for corruption, β^=0.4873\hat{\beta} = 0.4873 (z=10.51z = 10.51).

59
New cards

When re-estimating the parking violations model with Poisson PML (robust standard errors), what happened to the z-statistics?

The coefficient estimates remained identical, but the robust standard errors increased, causing the zz-statistic for staff to fall to 2.232.23 (p=0.026p = 0.026) and for corruption to fall to 2.672.67 (p=0.008p = 0.008).

60
New cards

In the Negative Binomial estimation of parking tickets, what was the estimated overdispersion parameter alpha and its significance test?

α^=1.5840\hat{\alpha} = 1.5840 (Std. Err.=0.2007\text{Std. Err.} = 0.2007); the likelihood ratio test of α=0\alpha = 0 gave χˉ2=1772.10\bar{\chi}^2 = 1772.10 (p<0.001p < 0.001), confirming severe overdispersion.

61
New cards

What is the formula for the marginal effect of a continuous variable x2x_2 on E(y∣x)E(y \mid x) in count models?

∂E(y∣x)∂x2=β2exp⁡(β0+β1x1+β2x2)=β2E(y∣x)\frac{\partial E(y \mid x)}{\partial x_2} = \beta_2 \exp(\beta_0 + \beta_1 x_1 + \beta_2 x_2) = \beta_2 E(y \mid x).

62
New cards

How is the average marginal effect of a continuous variable x2x_2 approximated in count models?

AME≈β2yˉ\text{AME} \approx \beta_2 \bar{y}, where yˉ\bar{y} is the sample mean of the dependent count variable.

63
New cards

In the Poisson PML parking tickets model, what was the average marginal effect of staff?

AME=0.0164049×11.27517≈0.1858\text{AME} = 0.0164049 \times 11.27517 \approx 0.1858 (p=0.036p = 0.036), meaning one additional diplomat increases expected unpaid violations by approximately 0.190.19.

64
New cards

How is the coefficient βk\beta_k in a Poisson or Negative Binomial model interpreted as a relative effect?

As a semi-elasticity: βk=∂E(y∣x)/∂xkE(y∣x)\beta_k = \frac{\partial E(y \mid x)/\partial x_k}{E(y \mid x)}, meaning a 1-unit increase in xkx_k changes the expected count by approximately βk×100%\beta_k \times 100\%.

65
New cards

In the Poisson parking tickets model, what is the relative effect interpretation of β^staff=0.0164\hat{\beta}_{\text{staff}} = 0.0164?

One additional diplomatic staff member increases the expected number of parking violations by 1.64%1.64\%, holding other variables fixed.

66
New cards

When comparing fitted counts to actual parking violations, what were the squared correlations for Poisson PML and Negative Binomial?

For Poisson PML, r2=0.1007r^2 = 0.1007; for Negative Binomial, r2=0.0056r^2 = 0.0056.

67
New cards

According to the countfit results for parking violations, how did Poisson and NB compare in predicting the proportion of zero counts (actual = 0.189)?

The standard Poisson model severely underpredicted zeros at 0.0180.018, while the Negative Binomial model accurately predicted zeros at 0.1940.194.

68
New cards

What is the recommended workaround in Stata if countfit produces the error 'factor-variable and time-series operators not allowed'?

Manually generate dummy variables for each category level (e.g., tab year, gen(yeardummy)) and include the created dummies in the regression rather than using factor notation (e.g., i.year).

69
New cards

What is reporting heterogeneity (or cut-point shift) in self-reported survey data?

A phenomenon where different individuals or groups have different internal thresholds/standards (τj\tau_j) when mapping the same underlying latent state (y∗y^*) into categorical response options (yy).

70
New cards

What evidence from the European Community Household Panel (ECHP 2001) illustrated reporting heterogeneity in self-reported health?

Respondents in Denmark reported much higher health ratings than respondents in Portugal, even though objective epidemiological measures showed no such large health discrepancy.

71
New cards

What are several potential sources of reporting heterogeneity identified in the lecture?

Differences in standards, awareness, expectations, adaptation to chronic illness, reference groups, and cultural norms.

72
New cards

Why is the standard ordered probit model problematic when reporting heterogeneity is present?

It assumes constant cut-points across all individuals, causing differences in reporting standards to be falsely attributed to true differences in the latent index.

73
New cards

In the SHARE France breathing ability study, what was the dependent variable y?

Self-rated shortness of breath in the last 30 days, categorized into five ordered options: Extreme, Severe, Moderate, Mild, and None (N=837N = 837).

74
New cards

What was the result of testing the effect of education on breathing problems in the standard ordered probit model?

The joint Wald test on the education dummies gave χ2(3)=2.41\chi^2(3) = 2.41 (p=0.4918p = 0.4918), indicating no statistically significant effect of education on breathing ability.

75
New cards

What is the Hierarchical Ordered Probit (HOPIT) model?

An extended ordered probit model that parameterizes the cut-point thresholds τj\tau_j as linear functions of individual characteristics to account for response heterogeneity.

76
New cards

How does the HOPIT model specify the individual-specific cut-points τj\tau_j mathematically?

τj=γ0,j+γ1,jx1+γ2,jx2\tau_j = \gamma_{0,j} + \gamma_{1,j} x_1 + \gamma_{2,j} x_2 for j=1,…,J−1j = 1, \dots, J-1.

77
New cards

Why can the HOPIT model not be identified using self-assessed health data (y) alone?

Because without an external benchmark, shifts in the cut-points (γ\gamma) cannot be mathematically distinguished from shifts in the latent index (β\beta).

78
New cards

What are anchoring vignettes, and how do they resolve the identification problem in HOPIT models?

Hypothetical descriptions of fixed health states evaluated by all respondents; because the described state is identical, any systematic variation in vignette ratings reflects differing cut-points rather than differences in health.

79
New cards

What is the latent equation for a vignette rating vlv_l in the HOPIT model?

vl∗=αl+ulv_l^* = \alpha_l + u_l, where ul∼N(0,1)u_l \sim N(0,1) and αl\alpha_l is a constant fixed health state, meaning vl∗v_l^* is not a function of individual covariates.

80
New cards

What is the 'Vignette Equivalence' assumption in the HOPIT framework?

The assumption that all respondents perceive and interpret the hypothetical health state described in a vignette in the same way, up to a random i.i.d. error term.

81
New cards

What is the 'Response Consistency' assumption in the HOPIT framework?

The assumption that an individual applies the exact same threshold cut-points (τj\tau_j) when evaluating their own health as they do when evaluating anchoring vignettes.

82
New cards

In the French breathing study, what happened to the education coefficients on latent health (y∗y^*) in HOPIT compared to ordered probit?

Education became positive and highly significant (level 3: β^=0.3283\hat{\beta} = 0.3283, p=0.010p = 0.010; level 4: β^=0.4990\hat{\beta} = 0.4990, p=0.001p = 0.001), revealing that higher education increases true breathing ability.

83
New cards

What did the HOPIT cut-point estimates reveal about the reporting standards of higher-educated individuals in France?

Higher-educated individuals have significantly higher cut-points (τ3\tau_3 and τ4\tau_4), meaning they have stricter health standards and are less likely to report having 'no problem' for a given breathing state.

84
New cards

What was the result of testing for reporting heterogeneity by education across all cut-points in the French breathing HOPIT model?

The joint Wald test across the 12 education threshold coefficients yielded χ2(12)=31.85\chi^2(12) = 31.85 (p=0.0015p = 0.0015), rejecting the null of reporting homogeneity at the 1%1\% level.

85
New cards

What was the result of testing for reporting heterogeneity jointly across education, gender, and age in the HOPIT model?

χ2(20)=49.87\chi^2(20) = 49.87 (p=0.0002p = 0.0002), confirming that cut-points vary significantly across demographic characteristics.

86
New cards

Why are raw predicted probabilities P(y=j)P(y = j) not directly comparable across individuals in the presence of response heterogeneity?

Because each individual evaluates their latent state against a different personal set of threshold cut-points (τj\tau_j).

87
New cards

How are predicted probabilities 'cleaned' of reporting heterogeneity in the HOPIT model?

By calculating probabilities using the estimated latent index coefficients (β^\hat{\beta}) while fixing the cut-point thresholds (τj\tau_j) to those of a chosen reference individual for everyone in the sample.

88
New cards

What reference profile was used to fix cut-point τ4\tau_4 when evaluating average marginal effects on P(y=5)P(y=5) in the French breathing example?

A male individual, aged 64, with the lowest education level (educ1).

89
New cards

How did the marginal effect of highest education (educ4) on P(y=5)P(y=5) (no breathing problems) change after HOPIT correction?

It increased from an insignificant +0.0676+0.0676 in ordered probit to a statistically significant +0.1811+0.1811 (p<0.001p < 0.001) in HOPIT.

90
New cards

What was the major substantive conclusion reached by Bago d'Uva, O'Donnell, and van Doorslaer (2008)?

Education-related health inequalities are often severely underestimated or go undetected in uncorrected survey data because higher-educated individuals evaluate health with stricter standards; correcting for reporting heterogeneity reveals significant inequalities.