1.Microeconometric Modeling
William Greene
Stern School of Business
New York University
New York NY USA
Models for OrderedChoices
http://people.stern.nyu.edu/Econometrics/OrderedChoices.pptx
2.Ordered Discrete Outcomes
E.g.: Taste test, credit rating, course grade, preference scale
Underlying random preferences:
Existence of an underlying continuous preference scale
Mapping to observed choices
Strength of preferences is reflected in the discrete outcome
Censoring and discrete measurement
The nature of ordered data
3.Ordered Choices at IMDb
4.
5.
6.
7.This study analyzes ‘self assessed health’ coded
1,2,3,4,5 = very low, low, med, high very high
8.Health Satisfaction (HSAT)
Self administered survey: Health Care Satisfaction (0 – 10)
Continuous Preference Scale
9.Modeling Ordered Choices
Random Utility (allowing a panel data setting)
Uit = + ’xit + it
= ait + it
Observe outcome j if utility is in region j
Probability of outcome = probability of cell
Pr[Yit=j] = F(j – ait) - F(j-1 – ait)
10.Ordered Probability Model
11.Combined Outcomes for Health Satisfaction
(0,1,2) (3,4,5) (6,7,8) (9) (10)
12.Ordered Probabilities
13.An Ordered Probability Model for Health Satisfaction
14.
15.Analysis of Model Implications
Partial Effects
Fit Measures
Predicted Probabilities
Averaged: They match sample proportions.
By observation
Segments of the sample
Related to particular variables
16.Coefficients
17.Partial Effects of 8 Years of Education
18.Ordered Probability Partial Effects
-----------------------------------------------------------------------------
Marginal effects for ordered probability model
M.E.s for dummy variables are Pr[y|x=1]-Pr[y|x=0]
Names for dummy variables are marked by *.
--------+--------------------------------------------------------------------
| Partial Prob. 95% Confidence
HLTHSAT| Effect Elasticity z |z|>Z* Interval
--------+--------------------------------------------------------------------
|--------------[Partial effects on Prob[Y=00] at means]--------------
*FEMALE| -.00117 -.02600 -.38 .7065 -.00726 .00492
EDUC| -.00351*** -.89008 -5.04 .0000 -.00488 -.00215
AGE| .00177*** 1.70456 11.15 .0000 .00146 .00208
INCOME| -.02298** -.17806 -2.37 .0178 -.04199 -.00398
*HHKIDS| -.00472 -.10470 -1.42 .1545 -.01121 .00177
|--------------[Partial effects on Prob[Y=01] at means]--------------
...
|--------------[Partial effects on Prob[Y=02] at means]--------------
...
|--------------[Partial effects on Prob[Y=03] at means]--------------
*FEMALE| .00146 .01323 .38 .7067 -.00614 .00906
EDUC| .00437*** .45292 4.82 .0000 .00259 .00615
AGE| -.00220*** -.86738 -9.36 .0000 -.00266 -.00174
INCOME| .02863** .09061 2.35 .0189 .00473 .05254
*HHKIDS| .00594 .05386 1.40 .1607 -.00236 .01424
|--------------[Partial effects on Prob[Y=04] at means]--------------
*FEMALE| .00192 .02209 .38 .7067 -.00808 .01191
EDUC| .00575*** .75573 5.05 .0000 .00352 .00798
AGE| -.00289*** -1.44727 -11.11 .0000 -.00341 -.00238
INCOME| .03764** .15118 2.37 .0178 .00651 .06878
*HHKIDS| .00786 .09053 1.40 .1618 -.00315 .01888
--------+--------------------------------------------------------------------
z, prob values and confidence intervals are given for the partial effect
***, **, * ==> Significance at 1%, 5%, 10% level.
19.-----------------------------------------------------------------------------
Marginal effects for ordered probability model
M.E.s for dummy variables are Pr[y|x=1]-Pr[y|x=0]
Names for dummy variables are marked by *.
[Partial effects on Prob[Y=j] at means]
--------+--------------------------------------------------------------------
| Partial Prob. 95% Confidence
HLTHSAT| Effect Elasticity z |z|>Z* Interval
--------+--------------------------------------------------------------------
*FEMALE| -.00117 -.02600 -.38 .7065 -.00726 .00492
*FEMALE| -.00304 -.01232 -.38 .7066 -.01890 .01281
*FEMALE| .00084 .00164 .38 .7065 -.00352 .00520
*FEMALE| .00146 .01323 .38 .7067 -.00614 .00906
*FEMALE| .00192 .02209 .38 .7067 -.00808 .01191
---------------------------------------------------------------------
Partial Effects Analysis for Ordered Probit Prob[Y =All]
Effects on function with respect to FEMALE
Results are computed by average over sample observations
Partial effects for binary var FEMALE computed by first difference
---------------------------------------------------------------------
df/dFEMALE Partial Standard
(Delta Method) Effect Error |t| 95% Confidence Interval
---------------------------------------------------------------------
APE Prob(y= 0) -.00124 .00329 .38 -.00768 .00521
APE Prob(y= 1) -.00288 .00765 .38 -.01788 .01212
APE Prob(y= 2) .00077 .00204 .38 -.00323 .00477
APE Prob(y= 3) .00138 .00367 .38 -.00581 .00857
APE Prob(y= 4) .00197 .00524 .38 -.00829 .01223
Partial Effects at Means vs. Average Partial Effects
20.Predictions from the Model Related to Age
21.Fit Measures
There is no single “dependent variable” to explain.
There is no sum of squares or other measure of “variation” to explain.
Predictions of the model relate to a set of J+1 probabilities, not a single variable.
How to explain fit?
Based on the underlying regression
Based on the likelihood function
Based on prediction of the outcome variable
22.Log Likelihood Based Fit Measures
23.
24.A Somewhat Better Fit
25.Panel Data
Fixed Effects
The usual incidental parameters problem
Partitioning Prob(yit > j|xit) produces estimable binomial logit models. (Find a way to combine multiple estimates of the same β.
Random Effects
Standard application
Extension to random parameters
Dynamics
Attrition
26.A Study of Health Status in the Presence of Attrition
27.Model for Self Assessed Health
British Household Panel Survey (BHPS)
Waves 1-8, 1991-1998
Self assessed health on 0,1,2,3,4 scale
Sociological and demographic covariates
Dynamics – inertia in reporting of top scale
Dynamic ordered probit model
Balanced panel – analyze dynamics
Unbalanced panel – examine attrition
28.Dynamic Ordered Probit Model
It would not be appropriate to include hi,t-1 itself in the model as this is a label, not a measure
29.Random Effects Dynamic Ordered Probit Model
30.Data
31.Variable of Interest
32.Dynamics
33.Attrition
34.Testing for Attrition Bias
Three variables added to full model with unbalanced panel suggest presence of attrition effects.
35.Estimated Partial Effects by Model
36.Partial Effect for a Category
These are 4 dummy variables for state in the previous period. Using first differences, the 0.234 estimated for SAHEX means transition from EXCELLENT in the previous period to GOOD in the previous period, where GOOD is the omitted category. Likewise for the other 3 previous state variables. The margin from ‘POOR’ to ‘GOOD’ was not interesting in the paper. The better margin would have been from EXCELLENT to POOR, which would have (EX,POOR) change from (1,0) to (0,1).
42.The Incidental Parameters Problem
Table 9.1 Monte Carlo Analysis of the Bias of the MLE in Fixed Effects Discrete Choice Models (Means of empirical sampling distributions, N = 1,000 individuals, R = 200 replications)
43.Zero Inflated Ordered Probit
44.Teenage Smoking
45.Inflated Responses in Self-Assessed Health
Mark Harris
Department of Economics, Curtin University
Bruce Hollingsworth
Department of Economics, Lancaster University
William Greene
Stern School of Business, New York University
46.SAH vs. Objective Health Measures
Favorable SAH categories seem artificially high.
60% of Australians are either overweight or obese (Dunstan et. al, 2001)
1 in 4 Australians has either diabetes or a condition of impaired glucose metabolism
Over 50% of the population has elevated cholesterol
Over 50% has at least 1 of the “deadly quartet” of health conditions (diabetes, obesity, high blood pressure, high cholestrol)
Nearly 4 out of 5 Australians have 1 or more long term health conditions (National Health Survey, Australian Bureau of Statistics 2006)
Australia ranked #1 in terms of obesity rates
Similar results appear to appear for other countries
47.A Two Class Latent Class Model
True Reporter
Misreporter
48.Mis-reporters choose either good or very good
The response is determined by a probit model
Y=3
Y=2