8:34 p.m.
Hi all,
I'm writing a script to implement the cointegration-with-break tests in Davidson &
Monticini (2010). The part I've done so far is what they call the
"incremental" version. Basically test for CI over subsamples, with the ending
date increasing by one each time.
The following block works fine in my main script:
<hansl>
# forward incremental
begdate = firstobs(y)
enddate = t0
loop while enddate<=n --quiet
smpl begdate enddate
ols y 0 time x --quiet
z = $uhat
coeffs[enddate,] = $coeff'
stderrs[enddate,] = $stderr'
adf 2 z --ct --test-down --quiet
stats[enddate] = $test
enddate++
endloop
</hansl>
But when I try to put this in a function, the smpl command generates an error:
Bad character 'b' in date string
error in new starting obs
*** error in function minmin
smpl begdate enddate
The variables 'begdate' & 'enddate' are getting set properly. I'm
using the shapshot built on 7/7, windows 7.
Any thoughts?
TIA,
PS
3:38 a.m.
On Wed, Jul 13, 2011 at 4:34 PM, Summers, Peter <psummers(a)highpoint.edu> wrote:
Hi all,
I'm writing a script to implement the cointegration-with-break tests in Davidson
& Monticini (2010). The part I've done so far is what they call the
"incremental" version. Basically test for CI over subsamples, with the ending
date increasing by one each time.
The following block works fine in my main script:
<hansl>
# forward incremental
begdate = firstobs(y)
enddate = t0
loop while enddate<=n --quiet
smpl begdate enddate
ols y 0 time x --quiet
z = $uhat
coeffs[enddate,] = $coeff'
stderrs[enddate,] = $stderr'
adf 2 z --ct --test-down --quiet
stats[enddate] = $test
enddate++
endloop
</hansl>
But when I try to put this in a function, the smpl command generates an error:
Bad character 'b' in date string
error in new starting obs
*** error in function minmin
> smpl begdate enddate
I think we need more context here. Perhaps you're trying to set the sample
range, inside the function, outside of the limits of the data as given to the
function?
E.g., if I do "smpl 1990:1 2010:4" before invoking a function, and then
in the function we have
smpl 1989:1 ;
that's an error -- even if the dataset goes back to, say, 1980:1.
Allin Cottrell
12:13 p.m.
<snip>
I think we need more context here. Perhaps you're trying to set
the sample
range, inside the function, outside of the limits of the data as given to the
function?
E.g., if I do "smpl 1990:1 2010:4" before invoking a
function, and then
in the function we have
smpl 1989:1 ;
that's an error -- even if the dataset goes back to, say, 1980:1.
I don't think that's it. The main script sets smpl 1950 2005, then the function
sets the first sample at 1950-1977, then increments the ending date by one each time
through the loop (up to 2005). There's also a backward part that starts from 1977-2005
and decrements the beginning date down to 1950.
PS
3:04 a.m.
On Thu, Jul 14, 2011 at 8:13 AM, Summers, Peter <psummers(a)highpoint.edu> wrote:
<snip>
>I think we need more context here. Perhaps you're trying to set the sample
>range, inside the function, outside of the limits of the data as given to the
>function?
>E.g., if I do "smpl 1990:1 2010:4" before invoking a function, and then
>in the function we have
>smpl 1989:1 ;
>that's an error -- even if the dataset goes back to, say, 1980:1.
I don't think that's it. The main script sets smpl 1950 2005, then the function
sets the first sample at 1950-1977, then increments the ending date by one each time
through the loop (up to 2005). There's also a backward part that starts from 1977-2005
and decrements the beginning date down to 1950.
OK, but we still need more context -- ideally, a minimal full script that
illustrates the problem.
Allin Cottrell
12:08 p.m.
________________________________________
From: gretl-users-bounces(a)lists.wfu.edu [gretl-users-bounces(a)lists.wfu.edu] on behalf of
Cottrell, Allin [cottrell(a)wfu.edu]
Sent: Thursday, July 14, 2011 11:04 PM
To: Gretl list
Subject: Re: [Gretl-users] smpl in a function
On Thu, Jul 14, 2011 at 8:13 AM, Summers, Peter <psummers(a)highpoint.edu> wrote:
<snip>
>I think we need more context here. Perhaps you're trying to set the sample
>range, inside the function, outside of the limits of the data as given to the
>function?
>E.g., if I do "smpl 1990:1 2010:4" before invoking a function, and then
>in the function we have
>smpl 1989:1 ;
>that's an error -- even if the dataset goes back to, say, 1980:1.
I don't think that's it. The main script sets smpl 1950 2005, then the function
sets the first sample at 1950-1977, then increments the ending date by one each time
through the loop (up to 2005). There's also a backward part that starts from 1977-2005
and decrements the beginning date down to 1950.
OK, but we still need more context -- ideally, a minimal full script
that
illustrates the problem.
Allin et al,
Attached is a copy of the script I've been working on, along with a data set. Once I
got the main program to work, I just copied the relevant parts into the function at the
top. Just uncomment the function call (and comment out the rest) to switch back &
forth.
PS
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
3:37 p.m.
New subject: omit --auto
Is there any way of saving in a list the significative variables that
results after the 'omit --auto' command?
I tried
<hansl>
ols Y const X1 X2 (...etc)
omit --auto
list uvar=$xlist
</hansl>
but I am seeing that $xlist maintains the variables from the initial
regression.
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
12:46 a.m.
New subject: omit --auto
On Fri, Jul 15, 2011 at 11:37 AM, Ignacio Diaz-Emparanza
<ignacio.diaz-emparanza(a)ehu.es> wrote:
Is there any way of saving in a list the significative variables
that
results after the 'omit --auto' command?
I tried
<hansl>
ols Y const X1 X2 (...etc)
omit --auto
list uvar=$xlist
</hansl>
but I am seeing that $xlist maintains the variables from the initial
regression.
Were any variables in fact omitted in this case? I just tried
<hansl>
open data4-10
ols 1 0 2 3 4 5 6
list XL1 = $xlist
list XL1 print
omit --auto
list XL2 = $xlist
list XL2 print
</hansl>
and got output for the two lists:
? XL1
const CATHOL PUPIL WHITE ADMEXP REV
? XL2
const CATHOL WHITE
Allin Cottrell
7:33 a.m.
New subject: omit --auto
El 16/07/11 02:46, Cottrell, Allin escribió:
On Fri, Jul 15, 2011 at 11:37 AM, Ignacio Diaz-Emparanza
<ignacio.diaz-emparanza(a)ehu.es> wrote:
> Is there any way of saving in a list the significative variables that
> results after the 'omit --auto' command?
>
> I tried
>
> <hansl>
> ols Y const X1 X2 (...etc)
> omit --auto
> list uvar=$xlist
> </hansl>
>
> but I am seeing that $xlist maintains the variables from the initial
> regression.
Were any variables in fact omitted in this case? I just tried
<hansl>
open data4-10
ols 1 0 2 3 4 5 6
list XL1 = $xlist
list XL1 print
omit --auto
list XL2 = $xlist
list XL2 print
</hansl>
and got output for the two lists:
? XL1
const CATHOL PUPIL WHITE ADMEXP REV
? XL2
const CATHOL WHITE
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
Yes there were omitted variables. I am seeing that I can repeat your
script obtaining your same results, but with my data the lists of
variables generadted before and after the 'omit' command is the same.
Here are my results:
<output>
? m1_ols <- ols renta2006 const capital l_regre l_regre2
m1_ols:
OLS, using observations 1-250 (n = 192)
Missing or incomplete observations dropped: 58
Dependent variable: renta2006
coefficient std. error t-ratio p-value
---------------------------------------------------------------
const 14639.2 3270.20 4.477 1.69e-05 ***
capital 224.908 1136.41 0.1979 0.8434
pibpc06 0.258699 0.190965 1.355 0.1780
GasT06_pc -7.69157 11.9757 -0.6423 0.5219
GasCo06_pc 8.53812 11.8751 0.7190 0.4735
GasCa06_pc 7.92740 11.7618 0.6740 0.5016
EstIndu06_pc -9.56127 19.6812 -0.4858 0.6280
EstCons06_pc -193.165 116.651 -1.656 0.1002
EstCom06_pc 219.689 107.176 2.050 0.0425 **
EstHos06_pc -424.834 225.760 -1.882 0.0622 *
EstTrans06_pc 195.715 127.908 1.530 0.1285
EstInte06_pc -143.172 173.804 -0.8238 0.4116
EstOt06_pc 58.0083 98.6155 0.5882 0.5574
Emp_Ind06_pc -23587.9 16585.1 -1.422 0.1575
Emp_Con06_pc 5984.31 28232.8 0.2120 0.8325
Emp_Com06_pc -19324.2 14823.1 -1.304 0.1947
Emp_Hos06_pc 41439.1 15985.1 2.592 0.0107 **
Emp_Tra06_pc -47031.9 17412.0 -2.701 0.0079 ***
Emp_Fin06_pc 14045.7 33464.9 0.4197 0.6754
Emp_OtS06_pc -32297.9 23291.0 -1.387 0.1680
Turis06_pc 1552.10 2171.71 0.7147 0.4761
Carga06_pc -4249.79 10086.2 -0.4213 0.6742
Par06_pc -62827.9 42745.2 -1.470 0.1441
Vpibpc06 -0.491249 0.210508 -2.334 0.0212 **
Gpibpc06 -0.198664 0.219040 -0.9070 0.3662
VGasT06_pc 3.35718 12.4759 0.2691 0.7883
GGasT06_pc 8.61628 12.1692 0.7080 0.4802
VGasCo06_pc -6.22659 12.4701 -0.4993 0.6184
GGasCo06_pc -10.4315 12.2720 -0.8500 0.3969
VGasCa06_pc -3.75115 12.2763 -0.3056 0.7604
GGasCa06_pc -8.71743 11.9407 -0.7301 0.4667
VEstIndu06_pc 31.8290 34.6712 0.9180 0.3604
GEstIndu06_pc -9.36813 35.8057 -0.2616 0.7940
VEstCons06_pc 104.586 127.755 0.8186 0.4145
GEstCons06_pc 148.486 130.305 1.140 0.2567
VEstCom06_pc -439.383 120.983 -3.632 0.0004 ***
GEstCom06_pc -209.136 125.003 -1.673 0.0968 *
VEstHos06_pc 669.718 262.448 2.552 0.0119 **
GEstHos06_pc 292.417 285.704 1.023 0.3080
VEstTrans06_pc -250.131 141.168 -1.772 0.0789 *
GEstTrans06_pc -264.133 156.396 -1.689 0.0937 *
VEstInte06_pc 477.038 177.869 2.682 0.0083 ***
GEstInte06_pc 188.988 189.803 0.9957 0.3213
VEstOt06_pc -144.880 118.941 -1.218 0.2255
GEstOt06_pc 66.9161 132.162 0.5063 0.6135
VEmp_Ind06_pc 42232.4 18274.3 2.311 0.0225 **
GEmp_Ind06_pc 20613.5 19048.4 1.082 0.2813
VEmp_Con06_pc 13433.7 28989.7 0.4634 0.6439
GEmp_Con06_pc -622.595 31859.7 -0.01954 0.9844
VEmp_Com06_pc 40153.2 16791.7 2.391 0.0183 **
GEmp_Com06_pc 19805.3 17162.1 1.154 0.2507
VEmp_Hos06_pc -9065.46 23261.2 -0.3897 0.6974
GEmp_Hos06_pc -55686.8 26272.7 -2.120 0.0360 **
VEmp_Tra06_pc 64215.9 18688.7 3.436 0.0008 ***
GEmp_Tra06_pc 41811.2 26335.8 1.588 0.1149
VEmp_Fin06_pc -28930.0 33943.7 -0.8523 0.3957
GEmp_Fin06_pc -13553.8 37773.6 -0.3588 0.7203
VEmp_OtS06_pc 60313.6 25886.5 2.330 0.0214 **
GEmp_OtS06_pc 25153.2 26955.1 0.9332 0.3525
VTuris06_pc -1845.25 2222.83 -0.8301 0.4080
GTuris06_pc -1335.91 2177.28 -0.6136 0.5406
VCarga06_pc 7311.69 10319.5 0.7085 0.4799
GCarga06_pc 3156.57 10110.4 0.3122 0.7554
VPar06_pc 43704.9 48815.4 0.8953 0.3723
GPar06_pc 54771.8 49922.0 1.097 0.2747
VConst 2431.35 3452.94 0.7041 0.4827
GConst -419.475 3655.47 -0.1148 0.9088
Mean dependent var 14207.78 S.D. dependent var 1869.337
Sum squared resid 2.41e+08 S.E. of regression 1387.303
R-squared 0.639551 Adjusted R-squared 0.449233
F(66, 125) 3.360446 P-value(F) 2.76e-09
Log-likelihood -1620.377 Akaike criterion 3374.754
Schwarz criterion 3593.007 Hannan-Quinn 3463.148
Excluding the constant, p-value was highest for variable 235 (GEmp_Con06_pc)
m1_ols saved
? list XL1 = $xlist
Generated list XL1
? list XL1 print
const capital pibpc06 GasT06_pc GasCo06_pc GasCa06_pc EstIndu06_pc \
EstCons06_pc EstCom06_pc EstHos06_pc EstTrans06_pc EstInte06_pc \
EstOt06_pc Emp_Ind06_pc Emp_Con06_pc Emp_Com06_pc Emp_Hos06_pc \
Emp_Tra06_pc Emp_Fin06_pc Emp_OtS06_pc Turis06_pc Carga06_pc Par06_pc \
Vpibpc06 Gpibpc06 VGasT06_pc GGasT06_pc VGasCo06_pc GGasCo06_pc \
VGasCa06_pc GGasCa06_pc VEstIndu06_pc GEstIndu06_pc VEstCons06_pc \
GEstCons06_pc VEstCom06_pc GEstCom06_pc VEstHos06_pc GEstHos06_pc \
VEstTrans06_pc GEstTrans06_pc VEstInte06_pc GEstInte06_pc VEstOt06_pc \
GEstOt06_pc VEmp_Ind06_pc GEmp_Ind06_pc VEmp_Con06_pc GEmp_Con06_pc \
VEmp_Com06_pc GEmp_Com06_pc VEmp_Hos06_pc GEmp_Hos06_pc VEmp_Tra06_pc \
GEmp_Tra06_pc VEmp_Fin06_pc GEmp_Fin06_pc VEmp_OtS06_pc GEmp_OtS06_pc \
VTuris06_pc GTuris06_pc VCarga06_pc GCarga06_pc VPar06_pc GPar06_pc \
VConst GConst
? omit --auto
Sequential elimination using two-sided alpha = 0.10
Dropping GEmp_Con06_pc (p-value 0.984)
Dropping GConst (p-value 0.895)
Dropping capital (p-value 0.823)
Dropping VGasT06_pc (p-value 0.804)
Dropping GEstIndu06_pc (p-value 0.834)
Dropping GCarga06_pc (p-value 0.747)
Dropping GEmp_Fin06_pc (p-value 0.713)
Dropping Emp_Fin06_pc (p-value 0.799)
Dropping VGasCa06_pc (p-value 0.568)
Dropping VEmp_Hos06_pc (p-value 0.605)
Dropping Emp_Con06_pc (p-value 0.588)
Dropping GEstOt06_pc (p-value 0.479)
Dropping GTuris06_pc (p-value 0.476)
Dropping EstInte06_pc (p-value 0.444)
Dropping VTuris06_pc (p-value 0.324)
Dropping Turis06_pc (p-value 0.478)
Dropping Carga06_pc (p-value 0.694)
Dropping EstOt06_pc (p-value 0.227)
Dropping VEstCons06_pc (p-value 0.248)
Dropping GEstCons06_pc (p-value 0.337)
Dropping GEmp_Com06_pc (p-value 0.445)
Dropping GEmp_OtS06_pc (p-value 0.533)
Dropping Gpibpc06 (p-value 0.751)
Dropping GEmp_Tra06_pc (p-value 0.438)
Dropping Emp_Com06_pc (p-value 0.321)
Dropping VConst (p-value 0.308)
Dropping VEstOt06_pc (p-value 0.438)
Dropping GPar06_pc (p-value 0.211)
Dropping GEstTrans06_pc (p-value 0.393)
Dropping EstTrans06_pc (p-value 0.726)
Dropping VPar06_pc (p-value 0.347)
Dropping VEstTrans06_pc (p-value 0.355)
Dropping GEstCom06_pc (p-value 0.282)
Dropping GEmp_Ind06_pc (p-value 0.279)
Dropping Par06_pc (p-value 0.141)
Dropping GEstHos06_pc (p-value 0.126)
Model 2: OLS, using observations 1-250 (n = 192)
Missing or incomplete observations dropped: 58
Dependent variable: renta2006
coefficient std. error t-ratio p-value
---------------------------------------------------------------
const 15135.2 517.080 29.27 2.34e-66 ***
pibpc06 0.0905194 0.0330375 2.740 0.0068 ***
GasT06_pc -7.08718 2.86912 -2.470 0.0145 **
GasCo06_pc 6.78625 3.00902 2.255 0.0255 **
GasCa06_pc 7.09263 2.87351 2.468 0.0146 **
EstIndu06_pc -20.3169 7.95165 -2.555 0.0115 **
EstCons06_pc -98.3470 21.5002 -4.574 9.49e-06 ***
EstCom06_pc 36.0169 17.2788 2.084 0.0387 **
EstHos06_pc -152.694 60.9441 -2.505 0.0132 **
Emp_Ind06_pc -7042.73 2985.53 -2.359 0.0195 **
Emp_Hos06_pc 33214.6 7775.64 4.272 3.31e-05 ***
Emp_Tra06_pc -13461.2 6337.52 -2.124 0.0352 **
Emp_OtS06_pc -12023.7 4034.46 -2.980 0.0033 ***
Vpibpc06 -0.330523 0.0836043 -3.953 0.0001 ***
GGasT06_pc 7.87246 3.26083 2.414 0.0169 **
VGasCo06_pc -1.56709 0.836233 -1.874 0.0627 *
GGasCo06_pc -8.21540 3.45846 -2.375 0.0187 **
GGasCa06_pc -7.56213 3.27700 -2.308 0.0223 **
VEstIndu06_pc 45.3765 25.9637 1.748 0.0824 *
VEstCom06_pc -253.867 50.2312 -5.054 1.16e-06 ***
VEstHos06_pc 404.973 119.907 3.377 0.0009 ***
VEstInte06_pc 337.274 37.2952 9.043 4.66e-16 ***
GEstInte06_pc 77.9347 41.3700 1.884 0.0614 *
VEmp_Ind06_pc 26611.5 7197.75 3.697 0.0003 ***
VEmp_Con06_pc 19243.0 6553.58 2.936 0.0038 ***
VEmp_Com06_pc 21961.8 7255.67 3.027 0.0029 ***
GEmp_Hos06_pc -45979.9 15967.2 -2.880 0.0045 ***
VEmp_Tra06_pc 28686.6 8682.05 3.304 0.0012 ***
VEmp_Fin06_pc -16798.0 4463.15 -3.764 0.0002 ***
VEmp_OtS06_pc 40182.0 10340.9 3.886 0.0001 ***
VCarga06_pc 1209.91 729.530 1.658 0.0992 *
Mean dependent var 14207.78 S.D. dependent var 1869.337
Sum squared resid 2.82e+08 S.E. of regression 1323.776
R-squared 0.577286 Adjusted R-squared 0.498520
F(30, 161) 7.329075 P-value(F) 7.76e-18
Log-likelihood -1635.674 Akaike criterion 3333.348
Schwarz criterion 3434.331 Hannan-Quinn 3374.247
Comparison of Model 1 and Model 2:
Null hypothesis: the regression parameters are zero for the variables
capital, EstTrans06_pc, EstInte06_pc, EstOt06_pc, Emp_Con06_pc,
Emp_Com06_pc, Emp_Fin06_pc, Turis06_pc, Carga06_pc, Par06_pc,
Gpibpc06, VGasT06_pc, VGasCa06_pc, GEstIndu06_pc, VEstCons06_pc,
GEstCons06_pc, GEstCom06_pc, GEstHos06_pc, VEstTrans06_pc,
GEstTrans06_pc,
VEstOt06_pc, GEstOt06_pc, GEmp_Ind06_pc, GEmp_Con06_pc, GEmp_Com06_pc,
VEmp_Hos06_pc, GEmp_Tra06_pc, GEmp_Fin06_pc, GEmp_OtS06_pc,
VTuris06_pc,
GTuris06_pc, GCarga06_pc, VPar06_pc, GPar06_pc, VConst, GConst
Test statistic: F(36, 125) = 0.599797, with p-value = 0.96105
Of the 3 model selection statistics, 3 have improved.
? list XL2 = $xlist
Generated list XL2
? list XL2 print
const capital pibpc06 GasT06_pc GasCo06_pc GasCa06_pc EstIndu06_pc \
EstCons06_pc EstCom06_pc EstHos06_pc EstTrans06_pc EstInte06_pc \
EstOt06_pc Emp_Ind06_pc Emp_Con06_pc Emp_Com06_pc Emp_Hos06_pc \
Emp_Tra06_pc Emp_Fin06_pc Emp_OtS06_pc Turis06_pc Carga06_pc Par06_pc \
Vpibpc06 Gpibpc06 VGasT06_pc GGasT06_pc VGasCo06_pc GGasCo06_pc \
VGasCa06_pc GGasCa06_pc VEstIndu06_pc GEstIndu06_pc VEstCons06_pc \
GEstCons06_pc VEstCom06_pc GEstCom06_pc VEstHos06_pc GEstHos06_pc \
VEstTrans06_pc GEstTrans06_pc VEstInte06_pc GEstInte06_pc VEstOt06_pc \
GEstOt06_pc VEmp_Ind06_pc GEmp_Ind06_pc VEmp_Con06_pc GEmp_Con06_pc \
VEmp_Com06_pc GEmp_Com06_pc VEmp_Hos06_pc GEmp_Hos06_pc VEmp_Tra06_pc \
GEmp_Tra06_pc VEmp_Fin06_pc GEmp_Fin06_pc VEmp_OtS06_pc GEmp_OtS06_pc \
VTuris06_pc GTuris06_pc VCarga06_pc GCarga06_pc VPar06_pc GPar06_pc \
VConst GConst
? quit
Script done
</output>
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
12:22 a.m.
New subject: omit --auto
On Mon, Jul 18, 2011 at 3:33 AM, Ignacio Diaz-Emparanza
<ignacio.diaz-emparanza(a)ehu.es> wrote:
El 16/07/11 02:46, Cottrell, Allin escribió:
"Were any variables in fact omitted in this case? I just tried" [...]
Yes there were omitted variables. I am seeing that I can repeat your script
obtaining your same results, but with my data the lists of variables
generadted before and after the 'omit' command is the same. Here are my
results [...]
I've tried creating a dummy case with dozens of regressors, as in your
example, but it still gives me correct results -- I can't replicate the
problem of the $xlist after omission not reflecting the omissions. Can
you send me your script and data (ofllist, if the data are big)?
Thanks,
Allin
12:33 a.m.
On Fri, Jul 15, 2011 at 8:08 AM, Summers, Peter <psummers(a)highpoint.edu> wrote:
"Attached is a copy of the script I've been working on, along with a
data set. Once I got the main program to work, I just copied the
relevant parts into the function at the top. Just uncomment the
function call (and comment out the rest) to switch back & forth."
Thanks, Peter. The script as sent fails with the message
'realtbill' is not the name of a variable
And indeed, that name is undefined. I substituted "tbill" (the name of a series
in your data file) for "realtbill" and the script runs OK. The inline or
non-function version of the script gives a "teststat" value of -4.9584.
If I modify the script to run the function instead of the inline code
it also gives -4.9584.
I edited the script a little to make toggling between function and inline
a bit easier, and I've attached my version.
Allin Cottrell
8:09 p.m.
"Thanks, Peter. The script as sent fails with the message
'realtbill' is not the name of a variable"
My bad -- realtbill is in the full data set
"And indeed, that name is undefined. I substituted "tbill" (the name of a
series
in your data file) for "realtbill" and the script runs OK. The inline or
non-function version of the script gives a "teststat" value of -4.9584.
If I modify the script to run the function instead of the inline code
it also gives -4.9584.
I edited the script a little to make toggling between function and inline
a bit easier, and I've attached my version."
So you didn't get the error on the smpl command? I'm puzzled (though grateful that
it works now!) -- I was working with the full data set so realtbill was in there, and the
error was definitely being generated within the function. I'll try to reproduce it,
but in the meantime thanks much.
PS
8:02 p.m.
<snip>
"So you didn't get the error on the smpl command? I'm puzzled (though
grateful that it works now!) -- I was working with the full data set so realtbill was in
there, and the error was definitely being generated within the function. I'll try to
reproduce it, but in the meantime thanks much."
Bad news -- I reproduced the error. Good news -- I've tracked down the cause.
The data set I've been working on is annual, 1947-2008, which I'm restricting to
1950-2005. So in my main program I have
<hansl>
open annual
smpl 1950 2005
#
# other stuff...
#
list ylist = spot
loop foreach i ylist --quiet
teststats[i] = minmin($i,realtbill,"I",0.5)
endloop
</hansl>
Then the function "minmin" does forward & backward recursive testing. The
problem seems to be related to the smpl command in the main program (ie, the 2nd line
above). If I comment that out, the function works fine. Otherwise I get the error:
Bad character 'b' in date string
error in new starting obs
*** error in function minmin
smpl begdate enddate
I think this didn't happen for Allin when he tested it because the data file I posted
was already sub-sampled to 1950-2005, so the smpl command had no effect.
Things work as expected if I move the smpl command to the function:
<hansl>
function minmin(arguments)
smpl 1950 2005
#
# other stuff
#
end function
open annual
#
# other stuff...
#
list ylist = spot
loop foreach i ylist --quiet
teststats[i] = minmin($i,realtbill,"I",0.5)
endloop
</hansl>
So I've got things working, but I don't understand why the smpl in the original
main program should be generating that error.
PS
1:08 a.m.
Hi Peter,
Thanks for following this up. Here's the deal:
First, you got a misleading error message, "Bad character
'b' in date string". That is fixed in my source tree and
wll be fixed in CVS as soon as sourceforge CVS is up again
(unfortunately it's broken right now). However, the other
part of the error message, namely
error in new starting obs
was quite correct: your function tried to push the start of
the sample range outside of the range passed in by the
caller, which is not allowed.
This is because of the way that firstobs() and lastobs()
work at present -- as noted in the Function Reference.
That is, firstobs() gives the observation number for the
first non-missing value of its series argument, but this
may occur before the start of the current sample range.
There's a case for changing that, so that firstobs() and
lastobs() do not return values outside of the current
sample, but that would be backward incompatible so I'm
not going to make the change uniterally; I'll raise the
question on the devel list.
I'm attaching a modified (skeletal) version of your script
that shows how to deal with sampling inside a function.
If we decide to change firstobs() and lastobs() this will
still work, although the conditional stuff with the '?'
operator would become redundant.
Allin
4:41 p.m.
<snip>
"your function tried to push the start of
the sample range outside of the range passed in by the
caller, which is not allowed.
This is because of the way that firstobs() and lastobs()
work at present -- as noted in the Function Reference.
That is, firstobs() gives the observation number for the
first non-missing value of its series argument, but this
may occur before the start of the current sample range.
There's a case for changing that, so that firstobs() and
lastobs() do not return values outside of the current
sample, but that would be backward incompatible so I'm
not going to make the change uniterally; I'll raise the
question on the devel list.
"I'm attaching a modified (skeletal) version of your script
that shows how to deal with sampling inside a function.
If we decide to change firstobs() and lastobs() this will
still work, although the conditional stuff with the '?'
operator would become redundant."
Thanks, Allin. I'd finally figured out that firstobs() and lastobs() might not be what
I wanted, and modified the script along the lines you suggest (though your version is much
more elegant than mine!).
PS
5217
days inactive
5224
days old
13 comments
3 participants
participants (3)
-
Cottrell, Allin -
Ignacio Diaz-Emparanza -
Summers, Peter