How to Load Example Models#
In the tutorials and guides of the respy documentation, you will frequently encounter example models. These are pre-defined models that can be easily accessed with just one function call to facilitate an introductory workflow and use in explanatory material. The set of examples consists of simple as well as very advanced models to cover a wide range of use cases and challenges that come with an increasing degree of complexity in dynamic discrete choice models.
For instance, in the tutorials, you will encounter two very simple models that are based on the story of Robinson Crusoe. Other example models are based on actual economic publications and thus serve to illustrate the scalability of respy models.
You can access example models by typing
params, options, data = rp.get_example_model("model_name")
If you only want to get the params and options, set the argument with_data to False.
params, options = rp.get_example_model("model_name", with_data=False)
For more information, checkout the function in theAPI
Below are the example models that are currently available.
The package provides users with two toy models. These models can be used to acquaintance oneself with respy’s functionalities and can be encountered in the tutorials.
Check out the Tutorials
The models are called:
robinson_crusoe_basic
robinson_crusoe_extended
The models are centered around Robinson Crusoe, who is stranded on a desert island. In each period, Robinson decides between fishing or relaxing in a hammock. In the extended model, he might additionally ask for Friday’s advice to further develop his fishing skills.
These models are excellent examples to use for learning and prototyping: They include a small number of available choices and a low number of periods such that they are computationally feasible.
Overview of model characteristics defined by params
and options
:
Parameters |
|
---|---|
Number of choices |
2 |
Work choices |
Fishing |
Education choices |
None |
Leisure choices |
Hammock |
Number of parameters |
7 |
Shock Correlations |
Negative between fishing and hammock |
Options |
|
---|---|
Number of periods |
5 |
Solution draws |
100 |
Estimation draws |
100 |
Solution seed |
456 |
Simulation seed |
132 |
Estimation seed |
100 |
Estimation tau |
0.001 |
Parameters |
|
---|---|
Number of choices |
3 |
Work choices |
Fishing |
Education choices |
Friday |
Leisure choices |
Hammock |
Number of parameters |
15 |
Shock Correlations |
None |
Lagged choices |
Hammock period 1 |
Covariates |
Break in fishing, contemplation with Friday |
Options |
|
---|---|
Number of periods |
10 |
Simulation agents |
1000 |
Solution draws |
500 |
Estimation draws |
200 |
Solution seed |
456 |
Simulation seed |
132 |
Estimation seed |
500 |
Estimation tau |
0.001 |
Aside from toy models, respy also provides several models that stem from the economic literature on dynamic life-cycle models. The most simple examples are a group of models based on the following publication:
Keane, M. P., & Wolpin, K. I. (1994). The Solution and Estimation of Discrete Choice Dynamic Programming Models by Simulation and Interpolation: Monte Carlo Evidence. The Review of Economics and Statistics, 648-672.
In the study, Keane and Wolpin (1994) develop an approximate solution method which consists of Monte Carlo integration with simulation and an interpolation approach to ease the computational burden of solving the DCDP model. They utilize one model with three different parametrizations to assess their solution method. This model and its three parametrizations are used as example models in the respy interface.
They are called:
kw_94_one
kw_94_two
kw_94_three
The model consists of four mutually exclusive alternatives that individuals can choose in each period. Agents can either choose to work in one of two sectors a or b, invest in education or stay home. The work alternatives award a wage and experience, while school only awards experience.
Overview of model characteristics defined by params
and options
:
Parameters
kw_94_one
kw_94_two
kw_94_three
Number of choices
4
Work choices
Cccupation sector a, Cccupation sector b
Education choices
education
Either choices
home
Number of parameters
30
Initial schooling
10 periods
Maximal schooling
20 periods
Shock Correlations
None
None
Positive (a and b), negative (home and educ)
Lagged choices
Education in period 1
Covariates
Squared experience, break education, HS Degree
Options
Number of periods
40
Simulation agents
1000
Solution draws
500
Estimation draws
200
Solution seed
15
Simulation seed
132
Estimation seed
500
Estimation tau
0.001
Monte Carlo Sequence
random
A more advanced group of examples are given by the models developed by Keane and Wolpin (1997). In this study, the authors implement an empirical structural life-cycle model of occupational choice and human capital investment. They estimate their model on data from the National Longitudinal Survey of Youth (NLSY). The study includes a “basic” model parametrization that is very similar to the model of Keane and Wolpin (1994) and and “extended” parametrization that improves on the empirical fit of the basic model.
Keane, M. P., & Wolpin, K. I. (1997). The Career Decisions of Young Men. Journal of Political Economy, 105(3), 473-522.
respy supports both the basic and extended parametrization from the paper. They models are named:
kw_97_basic
kw_97_extended
However, the parametrization from the paper returns different life-cycle patterns for
respy than presented in the paper. You can thus also access our estimates based for
the models that are based on the same empirical data by adding _respy
to the model
name.
The models consist of three occupational choices (white collar, blue collar, and military), one educational choice (school), and a home option. Both models consider a life-cycle of 50 periods. These models are decidedly larger than the toy models and require a considerable amount of computation power to solve, especially the extended model.
Overview of model characteristics defined by params
and options
:
Parameters |
|
---|---|
Number of choices |
5 |
Work hoices |
Blue collar, White collar, Military |
Education choices |
School |
Either choices |
Home |
Number of parameters |
63 |
Initial schooling |
7-11 periods |
Maximal schooling |
20 periods |
Lagged choices |
None |
Covariates |
School degrees, squared experience |
Unobserved Heterogeneity |
4 types |
Options |
|
---|---|
Number of periods |
50 |
Simulation agents |
5000 |
Solution draws |
200 |
Estimation draws |
200 |
Solution seed |
456 |
Simulation seed |
132 |
Estimation seed |
500 |
Estimation tau |
500 |
Monte Carlo Sequence |
random |
Parameters |
|
---|---|
Number of choices |
5 |
Work choices |
Blue collar, White collar, Military |
Education choices |
Education |
Either choices |
Home |
Number of parameters |
115 |
Initial schooling |
7-11 periods |
Maximal schooling |
20 periods |
Lagged choices |
School or Home in period 1 |
Covariates |
School degrees, squared experience, age, any or no prev. experience, military dropout, break in schooling |
Unobserved Heterogeneity |
4 types |
Measurement Error Wage |
Yes |
Options |
Value |
---|---|
Number of periods |
50 |
Simulation agents |
5000 |
Solution draws |
200 |
Estimation draws |
200 |
Solution seed |
1 |
Simulation seed |
2 |
Estimation seed |
3 |
Estimation tau |
500 |
Monte Carlo Sequence |
random |
Another example model provided in the respy interface is the model presented in Keane and Wolpin (2000). The model incorporates an observable characteristic to account for race, aiming to analyze the effects of monetary incentive schemes designed to reduce racial differences in school attainment and earnings.
Keane, M. P., & Wolpin, K. I. (2000). Eliminating Race Differences in School Attainment and Labor Market Success. Journal of Labor Economics, 18(4), 614-652.
The model is named
kw_2000
The model is very similar to the extended model specification in Keane and Wolpin
(1997). Overview of model characteristics defined by params
and options
:
Parameters
Number of choices
5
Choices
home, school, blue collar, white collar, military
Work choices
blue collar, white collar, military
Education choices
school
Either choices
home
Number of parameters
125
Initial education
7-11 periods
Maximal Schooling
20 periods
Correlations
positive correlation for all work alternatives
Lagged choices
School or Home in Period 1
Covariates
School degrees, squared experience, age, any or no prev. experience, military dropout, break in schooling
Observables
Ethnicity
Measurement Error Wage
Yes
Options
Number of periods
50
Simulation agents
5000
Solution draws
500
Estimation draws
200
Solution seed
456
Simulation seed
132
Estimation seed
500
Estimation Tau
500