Readme file
SERIES C
 Applied Statistics
A statistical model for the temporal pattern of individual automated teller machine withdrawals, by A. R. Brentnall, M. J. Crowder and D. J. Hand, pages 43–59;
--------------
Data Files
--------------
No data files are provided for confidentiality reasons.
-------------------
Program Files
-------------------
1. SOFTWARE
The estimation and prediction routines were written for GNU Octave 2.1.57 . Octave is freely available under the GNU Public License (GPL) from <http://www.gnu.org/software/octave/>.
The maximum likelihood estimation routines use a function called <nmsmax>, a Nelder-Mead simplex algorithm. If this is not part of your Octave distribution then you may obtain it from <http://octave.sourceforge.net/>.
2. FUNCTIONS
The m-file functions for estimation and prediction are found in the directory <Functions/>. A detailed description of each function, including input and output, is provided in each file. A summary of six function groups follows.
A.
ATMPP_GenVarTrapTimes.m
- This function generates time points to be used by a trapezium rule for numerical integration.
B.
ATMPP_FP_MLE.m
- This function is called to estimate the maximum-likelihood parameter set for an individual.
C.
ATMPP_FP_LOGLIK.m
ATMPP_LIK_FP_1.m
ATMPP_LIK_FP_2.m
- These are used to calculate the log-likelihood.
D.
ATMPP_P1.m
ATMPP_P1_2.m
ATMPP_P2.m
ATMPP_P2_1.m
ATMPP_F3.m
ATMPP_F3_2.m
- These are used to select the form of the rate function. Only one type of rate function is available, but the routines might be extended if other rate functions are preferred.
E.
ATMPP_Posterior.m
ATMPP_UPost.m
ATMPP_PredictZero.m
ATMPP_Predict.m
- These routines are used in prediction.
F.
ATMPP_Monitor.m
ATMPP_RATE.m
- These functions may be used to monitor the cumulative rate function.
3. HOW TO FIT AND PREDICT
Example scripts on how to use the functions to fit and predict are provided in the main directory. See <Fitdata.m> for fitting, and <PredictFuture.m> for monitoring and prediction.
Adam Brentnall
Institute for Mathematical Sciences
Imperial College London
53 Princes Gate
South Kensington
London
SW7 2PG
UK
E-mail: a.brentnall@imperial.ac.uk
|
