Paul - dropping the constant definitely helps. I'm now able to estimate
a reasonable model. As for the households for estimation data, we're
using some household level market segmentation data that was purchased
from Claritas. This data set includes some of the basic (age of head,
income, presence of children) information that we're likely to need for
estimation. This is different from our synthesized households.
Unfortunately the data does not indicate how long households have lived
in their current location, but based on my observation seems better to
use than the synthetic households because the relationship between age,
etc and the household location is more accurate than the information
generated in the household synthesis and assignment process.
In reading your 2006 paper "Reconciling household residential location
choices ..." you mention the use of household level surveys. Was this
part of a trip diary survey or simply a household level survey targeted
at recent movers? Is it possible to see the survey questionnaire that
was used to collect this data? I'm hoping that we can collect more
specific household level data in the near future.
brian
Paul Waddell wrote:
> Brian
>
> It looks like you're including a constant, which would not be
> appropriate in a location choice model where the model estimation is
> using random sampling of alternatives. You can only use alternative-
> specific constants when you have a small number of alternatives, and
> they are always in the same order-- like in a mode choice model. Try
> dropping the constant for starters.
>
> Also - I can't tell if you have a survey of households with observed
> locations, or are sampling from synthetic households. The former is
> definitely better, especially if you can sample only those who have
> moved within the past few years.
>
> Paul
>
> On Feb 29, 2008, at 11:12 AM, brian voigt wrote:
>
>
>> I'm currently working on estimating the Household Location Choice
>> Model and am having trouble producing a worthwhile result. I'm not
>> able to get the model to converge no matter the combination of
>> variables that I have included in the estimation procedure.
>> Additionally, the convergence statistic is always the same as the
>> number of observations (even when I change the number of households
>> represented in the households_for_estimation table). Finally, as can
>> be seen in the output, the coefficient estimates are all extremely
>> small (with correspondingly small t-values). Does this indicate that
>> there is a problem with my households_for_estimation table?
>>
>> Perhaps I am confused about the function of the estimation
>> procedure. I understand that the number of movers and the unplaced
>> agents are derived by the Household Relocation Model, but how do
>> these movers and unplaced agents relate to the records in the
>> households_for_estimation table? More specifically, how can there be
>> more movers / unplaced agents than there are records in the
>> households_for_estimation table? Perhaps too I don't understand
>> correctly what the dependent variable is that we are modeling. I
>> thought the estimation procedure evaluated the location choice for
>> each household in the household_for_estimation table compared to the
>> attributes of the other records in the choice set. Is this correct?
>>
>> I've attached the output from an attempt at estimation to help
>> illustrate my point. Any input / advice would be most appreciated?
>> brian
>>
>> Brian Voigt
>> Spatial Analysis Lab
>> University of Vermont
>> Burlington, VT 05405
>>
>> brian.voigt_at_uvm.edu
>>
>> C:\myworkspace\chittenden\estimation>python run_estimation.py -c
>> chittenden.estimation.estimation_config_voigt --use-
>> trapperkeeper=false
>> Cache Directory set to: C://urbansim_cache/chittco_1990_baseyear_cache
>> Start simulation run: started on Fri Feb 29 13:18:32 2008
>> random seed = (1,)
>> Starting simulation for year 1990: started on Fri Feb 29 13:18:32
>> 2008
>> Closing log file: C://urbansim_cache/
>> chittco_1990_baseyear_cache\run_multiprocess.log
>> Logging to file: C://urbansim_cache/
>> chittco_1990_baseyear_cache\year_1990_log.txt
>> Simulate year 1990: started on Fri Feb 29 13:18:32 2008
>> Creating object
>> urbansim
>> .datasets
>> .vacant_land_and_building_type_dataset
>> .VacantLandAndBuildingTypeDataset
>> Creating object
>> urbansim
>> .datasets.development_constraint_dataset.DevelopmentConstraintDataset
>> Creating object urbansim.datasets.zone_dataset.ZoneDataset
>> Creating object
>> urbansim.datasets.household_dataset.HouseholdDataset
>> Creating object
>> urbansim.datasets.gridcell_dataset.GridcellDataset
>> Creating object urbansim.datasets.job_dataset.JobDataset
>> Creating object
>> urbansim.datasets.target_vacancy_dataset.TargetVacancyDataset
>> Creating object
>> urbansim.datasets.job_building_type_dataset.JobBuildingTypeDataset
>> Creating object
>> urbansim
>> .datasets
>> .development_event_history_dataset.DevelopmentEventHistoryDataset
>> Creating object urbansim.datasets.rate_dataset.RateDataset
>> Running Household Relocation Model (from
>> urbansim.models.agent_relocation_model): started on Fri Feb 29
>> 13:18:33 2008
>> compute_probabilities ...
>> compute_choices ...
>> Number of movers: 11301
>> Running Household Relocation Model (from
>> urbansim.models.agent_relocation_model): completed...4.8 sec
>> Unplace 941 agents.
>> Estimating Household Location Choice Model (from
>> urbansim.models.household_location_choice_model): started on Fri Feb
>> 29 13:18:38 200
>>
>> Number of agents for estimation: 5000
>> Sampling locations for estimation ...
>> Sampling done in 1 chunk(s).
>> Choice set size: 30
>> Compute variables ...
>>
>> urbansim.gridcell.is_in_floodplain.......................0.3 sec
>>
>> urbansim.gridcell.ln_distance_to_highway.................0.3 sec
>>
>> urbansim.gridcell.ln_residential_units_within_walking_distance
>>
>> urbansim.gridcell.residential_units_within_walking_distance....0.6 sec
>>
>> urbansim.gridcell.ln_residential_units_within_walking_distance:
>> completed...0.7 sec
>>
>> urbansim.gridcell.ln_total_employment_within_walking_distance
>>
>> urbansim.gridcell.total_employment_within_walking_distance
>>
>> urbansim.gridcell.number_of_jobs.................0.3 sec
>>
>> urbansim.gridcell.total_employment_within_walking_distance:
>> completed...0.3 sec
>>
>> urbansim.gridcell.ln_total_employment_within_walking_distance:
>> completed...0.3 sec
>>
>> urbansim.gridcell.ln_total_population_within_walking_distance
>>
>> urbansim.gridcell.total_population_within_walking_distance
>>
>> urbansim.gridcell.population.....................0.0 sec
>>
>> urbansim.gridcell.total_population_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim.gridcell.ln_total_population_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim.gridcell.number_of_jobs_within_walking_distance....0.1 sec
>> urbansim.gridcell.percent_SSS_within_walking_distance
>>
>> urbansim
>> .gridcell.number_of_development_type_group_SSS_within_walking_distance
>>
>> urbansim.gridcell.is_in_development_type_group_SSS....0.7 sec
>>
>> urbansim
>> .gridcell
>> .number_of_development_type_group_SSS_within_walking_distance:
>> completed...0.8 sec
>>
>> urbansim.gridcell.percent_SSS_within_walking_distance: completed...
>> 0.9 sec
>>
>> urbansim
>> .gridcell.percent_high_income_households_within_walking_distance
>>
>> urbansim
>> .gridcell.number_of_high_income_households_within_walking_distance
>>
>> urbansim.gridcell.number_of_high_income_households
>>
>> urbansim.household.is_high_income............0.0 sec
>>
>> urbansim.gridcell.number_of_high_income_households: completed...0.0
>> sec
>>
>> urbansim
>> .gridcell.number_of_high_income_households_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim.gridcell.number_of_households_within_walking_distance
>>
>> urbansim.gridcell.number_of_households...........0.0 sec
>>
>> urbansim.gridcell.number_of_households_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim
>> .gridcell.percent_high_income_households_within_walking_distance:
>> completed...0.2 sec
>> urbansim.gridcell.percent_SSS_within_walking_distance
>>
>> urbansim
>> .gridcell.number_of_development_type_group_SSS_within_walking_distance
>>
>> urbansim.gridcell.is_in_development_type_group_SSS....0.0 sec
>>
>> urbansim
>> .gridcell
>> .number_of_development_type_group_SSS_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim.gridcell.percent_SSS_within_walking_distance: completed...
>> 0.1 sec
>> urbansim.gridcell.percent_SSS_within_walking_distance
>>
>> urbansim
>> .gridcell.number_of_development_type_group_SSS_within_walking_distance
>>
>> urbansim.gridcell.is_in_development_type_group_SSS....0.0 sec
>>
>> urbansim
>> .gridcell
>> .number_of_development_type_group_SSS_within_walking_distance:
>> completed...0.1 sec
>>
>> urbansim.gridcell.percent_SSS_within_walking_distance: completed...
>> 0.1 sec
>>
>> urbansim.household.income_category.......................0.0 sec
>>
>> urbansim.household.is_without_children...................0.0 sec
>> urbansim.household_x_gridcell.cost_to_income_ratio
>> urbansim.gridcell.total_annual_rent
>>
>> urbansim.gridcell.total_residential_value........0.6 sec
>> urbansim.gridcell.total_annual_rent:
>> completed.......0.6 sec
>> urbansim.household_x_gridcell.cost_to_income_ratio:
>> completed...0.6 sec
>>
>> urbansim.household_x_gridcell.income_and_year_built......0.3 sec
>> Estimate ...
>> submodel: -2
>> WARNING: Cannot find increase
>> Akaike's Information Criterion (AIC): 34047.9738166
>> Number of Iterations: 10
>> ***********************************************
>> Log-likelihood is: -17005.9869083
>> Null Log-likelihood is: -17005.9869083
>> Likelihood ratio index: 0.0
>> Adj. likelihood ratio index: -0.00105845077366
>> Number of observations: 5000
>> Suggested |t-value| > 2.91842306587
>> Convergence statistic is: 5000.0
>> -----------------------------------------------
>> Coeff_names estimate std err t-
>> values
>> D2LAKE -2.20663e-016 0.000496508
>> -4.4443e-013
>> constant 9.5983e+015 1.65882e+014
>> 57.8622
>> BFLOOD -1.31402e-013 0.242338
>> -5.42225e-013
>> BLDHW 8.11058e-015 0.0318991
>> 2.54257e-013
>> BLDUW -3.49379e-015 0.0801334
>> -4.35996e-014
>> LE_W 2.20441e-015 0.0158806
>> 1.38812e-013
>> BP_TW 3.0224e-015 0.0879053
>> 3.43825e-014
>> JOBS_WWD -3.68946e-019 8.16678e-006
>> -4.51764e-014
>> PCW -6.93212e-018 0.00267387
>> -2.59254e-015
>> BPHIW 1.86301e-017 0.00165672
>> 1.12451e-014
>> PIW 1.83701e-016 0.00412956
>> 4.44844e-014
>> PRW -1.81968e-017 0.00109088
>> -1.66809e-014
>> PWATER -5.58252e-016 0.00327898
>> -1.70252e-013
>> PWETLA -2.67156e-016 0.00240074
>> -1.1128e-013
>> INC_CAT 0.359024 1.84247e+013
>> 1.94861e-014
>> NO_KIDS -10.5912 1.78019e+014
>> -5.94949e-014
>> COST_INC_RAT 4.78316e-016 0.0382962
>> 1.24899e-014
>> INC_YRBLT 1.85108e-022 6.06333e-009
>> 3.0529e-014
>> ***********************************************
>> Elapsed time: 12.9514644907 seconds
>> Estimating Household Location Choice Model (from
>> urbansim.models.household_location_choice_model): completed...19.6 sec
>> Simulate year 1990:
>> completed...................................26.4 sec
>> Closing log file: C://urbansim_cache/
>> chittco_1990_baseyear_cache\year_1990_log.txt
>> Starting simulation for year 1990:
>> completed........................26.4 sec
>> Start simulation run:
>> completed.........................................26.4 sec
>> Closing log file: C://urbansim_cache/chittco_1990_baseyear_cache
>> \run_multiprocess.log_______________________________________________
>> Users mailing list
>> Users_at_urbansim.org
>> http://www.urbansim.org/mailman/listinfo/users
>>
>
> _______________________________________________
> Users mailing list
> Users_at_urbansim.org
> http://www.urbansim.org/mailman/listinfo/users
>
Received on Wed Mar 05 2008 - 12:54:20 PST
This archive was generated by hypermail 2.2.0 : Wed Mar 05 2008 - 12:54:23 PST