README
1 This directory contains test code for the functions of the 'tseries'
2 segment of the CCM math library. Each test code file has a sample
3 of the correct output appended to it. Input data files named in
4 the test code header, are in the 'data' sub-directory.
5
6 The 'data' subdirectory contains time-series model definition files.
7 The format of these definition files is specfied below.
8
9 The 'generators' subdirectory contains the source code of programs
10 that will generate the sequence of a specfied time series.
11
12
13 Note: If your machine is big-endian or it employs a floating point
14 format different from the IEEE 64 bit standard for doubles
15 you must regenerate some of the test data files in the 'data'
16 subdirectory. See the 'generators' sub-directory README file
17 for directions on this.
18
19
20 ----------------------------------------
21 Format of time series model data files
22 ----------------------------------------
23 Time series model definition files
24-------------------------------------------------------------------------
25 A. Ordinary ARMA model definition
26
27 nar nma (# autoregressive # moving average)
28 p1 lag1
29 - - ( parameter - lag list
30 - - AR then MA , # = nar + nma = np )
31 pN lagN
32
33 random_seed
34
35 B. FACTOR model definition
36
37 nfc nar nma ndif (# factors # autoregressive # moving average
38 #difference order )
39 p1 lag1
40 - - ( parameter - lag list
41 - - factor , AR , MA ; # = nfc + nar +nma = np )
42 pN lagN
43
44 random_seed
45
46--------------------------------------------------------------------------
47
48 In the 'data' subdirectory files with a name beginning with 'tsf'
49 refer to factor models, while files with a name 'ts' refer to
50 ordinary ARIMA models. The sufix describes the type of file, with
51
52 .dat files contain a model definition
53 ( used as input to generate a series)
54 .d files contain a model specification
55 ( definition sans parameter values
56 used as input to estimate a time series fit)
57 .b files contain the binary data of a time series
58 ( generated by 'garma' (ARIMA) modesl or by
59 'gfarma' (factor models) )
60 .er files contain the resuidal output from a
61 model estimation attempt (this is optional
62 output from the estimation programs and can
63 be used as input to the residual analysis
64 program tresid.c)
65
66 Note: It is instructive to run test estimations in which the
67 model specification (.d file) does not match the
68 definition file used to create the series!
69
70 The following list specifies the command line parameters and
71 prompt responses required to run the standard tests, whose
72 output is appended to the test source code. A prompt may
73 specify several parameters.
74
75
76 tdrfmod 20 data/tfs2.dat
77 tdrmod 20 data/ts1.dat
78 tevfmod 30 data/tfs0.dat
79 tevmod 30 data/ts2.dat
80 tfmest data/tfs0.d data/tfs0.b
81 prompt response sequence: s s s f q n
82 tfsread data/tfs0.b
83 tmest data/ts0.d data/ts0.b
84 prompt response sequence: s s s f f q n
85 tparma data/ts2.dat
86 tresid data/ts0.er
87 prompt responses: 20 -4 4 16 : n
88 tsany data/ts0.b
89 prompt response: 0 2 20
90 tsarma 20 data/ts2.dat
91 tsdiff 2 data/ts6.b
92 tsread data/ts0.b
93 txmean data/ts6.b
94