Commits

Author Commit Message Labels Comments Date
Yiliang Xiong
change dataset filenames
Yiliang Xiong
some minor changes for testing batch run
Yiliang Xiong
divide all the observations into $nb subsamples and estimate them separately
Yiliang Xiong
drop redundant variables run all the 108120214 observations consumes too much memory
Yiliang Xiong
some minor changes on filenames
Yiliang Xiong
also minor changes of filename
Yiliang Xiong
minor changes of filename
Yiliang Xiong
clear all before set memory
Yiliang Xiong
put the log files in a subdirectory
Yiliang Xiong
define global settings
Yiliang Xiong
back up the original observation order
Yiliang Xiong
predict the rankings based on choice probability conditional on one positive response (pc1)
Yiliang Xiong
mark observations with complete entries (1) observations with incomplete entries cannot be used for estimation (2) the missing entries are generated and used for prediction only
Yiliang Xiong
predict command can produce out-of-sample predictions
Yiliang Xiong
confirm that invlogit(xb) == pu0
Yiliang Xiong
drop dummy observations and no need to duplicate variables $nc times
Yiliang Xiong
minor changes to check convergence
Yiliang Xiong
automatically generate values for missing entries
Yiliang Xiong
minor changes of log and data files
Yiliang Xiong
batch run all the datasets
Yiliang Xiong
use nr and dfp algorithms
Yiliang Xiong
change the predicting process accordingly for the dummy alternative
Yiliang Xiong
create a dummy alternative for choice occasion with all negative responses I hope this trick can avoid overestimating positive response probability.
Yiliang Xiong
bugfix: remove sum()
Yiliang Xiong
bugfix: sum() is running sum, total() is overall sum
Yiliang Xiong
substitute the missing value with mean value and random variable
Yiliang Xiong
log the likelihood value in each iteration
Yiliang Xiong
bugfix: explicitly compare with the negative responses
Yiliang Xiong
force to encode missing as zero
Yiliang Xiong
bugfix: use capture instead of confirm to check if the global macro is defined
  1. Prev
  2. Next