Ahmad Abu Hammour
2002-Apr-19 04:48 UTC
[R] Durbin-Watson test in packages "car" and "lmtest"
Hi, P-values in Durbin-Watson test obtained through the use of functions available in packages "lmtest" and "car" are different. The difference is quite significant. function "dwtest" in "lmtest" is much faster than "burbinwatson" in "car". Actually, you can take a nap while the latter trying to calculated Durbin-Watson test. My question is which p-value is better? Thank you, Ahmad Abu Hammour -------------- next part -------------- An HTML attachment was scrubbed... URL: https://stat.ethz.ch/pipermail/r-help/attachments/20020419/274907c1/attachment.html
Torsten Hothorn
2002-Apr-19 08:49 UTC
[R] Durbin-Watson test in packages "car" and "lmtest"
> Hi, > P-values in Durbin-Watson test obtained through the use of > functions available in packages "lmtest" and "car" are different. The > difference is quite significant. function "dwtest" in "lmtest" is much > faster than "burbinwatson" in "car". Actually, you can take a nap while > the latter trying to calculated Durbin-Watson test. My question is which > p-value is better?The answer is essencially given in ?durbin.watson and ?dwtest. The latter states that The p value is computed using a Fortran version of the Applied Statistics Algorithm AS 153 by Farebrother (1980, 1984). This algorithm is called "pan" or "gradsol". For large sample sizes the algorithm might fail to compute the p value; in that case a warning is printed and an approximate p value will be given; this p value is computed using a normal approximation with mean and variance of the Durbin-Watson test statistic. while ?durbin.watson says simulate: if `TRUE' p-values will be estimated by bootstrapping. What is a "quite significant" difference for p-values? Looking at the example from ?durbin.watson gives: R> durbin.watson(lm(fconvict ~ tfr + partic + degrees + mconvict, data=Hartnagel)) lag Autocorrelation D-W Statistic p-value 1 0.688345 0.6168636 0 R> dwtest(fconvict ~ tfr + partic + degrees + mconvict, data=Hartnagel) Durbin-Watson test data: fconvict ~ tfr + partic + degrees + mconvict DW = 0.6169, p-value = 6.96e-09 which is fairly close, so you might give us more details (that is: a working example) to see what the "difference" is (that is: a bug in either function or a difference due to simulation error / bad approximation ...). Torsten> > Thank you, > Ahmad Abu Hammour >-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Ahmad Abu Hammour
2002-Apr-19 21:05 UTC
[R] Durbin-Watson test in packages "car" and "lmtest"
Hi Torsten,
Here is an example in which P-values have a significant difference. The data is
a sub sample of a larger one.
Basically the model I am using is as follows:
y_{t+2}= y_{t+1} + y_t + I(x_{t+1}-z_{t+1}-w_{t+1}) +e_{t+2}
> ydata
y z w x
[1,] 0.0101705342 -5.363636567 0.80677087 -2.425968
[2,] -0.0040963887 -4.930336567 -0.01515276 -1.852668
[3,] 0.0007996203 -2.795136567 0.28412361 -0.849268
[4,] 0.0231375722 -2.523036567 -0.33650002 -0.315968
[5,] 0.0576658597 -2.925936567 -0.68112365 0.060732
[6,] 0.0930807057 -2.283836567 -1.59374728 0.214032
[7,] 0.0904861271 -1.908436567 -3.15057091 0.347332
[8,] 0.0815802452 -3.063936567 -3.60669454 -0.175968
[9,] 0.0463679856 -3.947236567 -2.89361816 -1.012668
[10,] 0.0283314000 -3.528136567 -1.68974179 -1.522668
[11,] 0.0158693958 -2.890536567 -1.03456542 -1.629268
[12,] -0.0177988767 -2.165736567 -0.23008905 -1.532668
[13,] -0.0263232785 -2.400436567 -0.29991268 -1.369268
[14,] -0.0377695478 -3.243136567 -0.31953631 -0.539268
[15,] -0.0520095622 -3.241336567 0.22174006 0.120732
[16,] -0.0773138815 -2.765236567 -0.22328356 1.467332
[17,] -0.0804307590 -2.833236567 0.09209281 2.107332
[18,] -0.0446189151 -3.710136567 0.57246918 2.057332
[19,] -0.0439761920 -4.232636567 0.11764555 2.687332
[20,] -0.0508314286 -3.445136567 0.91252192 4.874032
[21,] -0.0373526376 -2.123836567 1.03269829 5.340732
[22,] -0.0363366045 -4.549336567 1.16137466 3.614032
[23,] -0.0119795294 -4.660636567 2.10495103 1.744032
[24,] -0.0271166745 0.940863433 2.73172741 5.450732
[25,] -0.0319279899 2.972463433 3.93230378 7.954032
[26,] -0.0285519284 2.601663433 4.25698015 9.447332
[27,] -0.0362890374 3.900363433 4.49975652 11.397332
[28,] -0.0193261619 2.440563433 4.24003289 7.057332
[29,] -0.0211604332 4.133063433 3.62460926 5.907332
[30,] -0.0393599621 4.764863433 3.78798563 6.740732
[31,] -0.0368496844 2.967363433 3.09016201 5.137332
[32,] -0.0111656400 2.374163433 2.32173838 1.820732
[33,] -0.0049543837 2.858663433 0.59471475 0.570732
[34,] -0.0085112660 3.900063433 -0.96020888 0.424032
[35,] 0.0002988086 3.956163433 -1.40353251 0.500732
[36,] -0.0021387277 2.276463433 -1.99135614 0.720732
[37,] -0.0143497002 2.857863433 -1.40157977 1.270732
[38,] -0.0459512047 3.589163433 -1.78060340 2.574032
[39,] -0.0620556482 4.248163433 -2.43422702 3.537332
[40,] -0.0665723041 3.225363433 -2.45305065 1.830732
[41,] -0.0990262190 2.638163433 -2.34577428 1.634032
[42,] -0.0993943457 2.160963433 -2.00509791 0.780732
[43,] -0.0896954664 1.651763433 -1.79992154 0.164032
[44,] -0.1047509912 2.106963433 -1.59274517 0.117332
[45,] -0.1188326548 3.230463433 -1.40576880 2.007332
[46,] -0.1012159404 2.484063433 -1.60469242 -0.209268
[47,] -0.0975617833 2.185163433 -1.23841605 -0.442668
[48,] -0.0859923186 1.244563433 -0.99983968 -0.495968
[49,] -0.0467949427 -0.185636567 -1.19776331 -1.649268
[50,] -0.0351160760 -0.510836567 -0.50778694 -0.569268
[51,] -0.0233943108 -0.356136567 -0.53781057 0.347332
[52,] -0.0080094290 0.005063433 -0.71733420 -0.569268
[53,] 0.0312603960 -0.149336567 -0.69205783 -0.349268
[54,] 0.0614944090 0.181063433 -0.76188145 0.237332
[55,] 0.0775709808 0.795863433 -0.58080508 1.110732
[56,] 0.0952235601 1.131763433 -0.38712871 1.907332
[57,] 0.1069545973 1.509763433 0.06534766 2.887332
[58,] 0.1136253179 1.785763433 0.64482403 3.450732
[59,] 0.1275201332 1.348463433 1.06650040 3.410732
[60,] 0.1387864450 0.581263433 1.05767677 3.444032
[61,] 0.1244438565 1.253563433 1.35555315 4.164032
[62,] 0.1344048716 0.334263433 0.69162952 4.844032
[63,] 0.1487305153 -0.634036567 0.37670589 4.010732
[64,] 0.1420063384 -0.159036567 1.22928226 3.190732
[65,] 0.1469999823 -0.707536567 1.67200000 1.200732> library(car)
Attaching package `car':
The following object(s) are masked from package:base :
dfbetas rstudent > library(lmtest)
>
lmy=lm(y[-c(1,2)]~y[-c(1,65)]+y[-c(64,65)]+I(x-z-w)[-c(64,65)]-1,data=as.ts(ydata))
> dwtest(lmy)
Durbin-Watson test
data: lmy
DW = 2.0077, p-value = 0.4404> durbin.watson(lmy)
lag Autocorrelation D-W Statistic p-value
1 -0.01370151 2.007701 0.92>
So, which P-value should I adopt 0.44 or 0.92?
Thank you for your help.
Ahmad Abu Hammour
----- Original Message -----
From: Torsten Hothorn
Sent: Friday, April 19, 2002 4:51 AM
To: Ahmad Abu Hammour
Cc: R-help at stat.math.ethz.ch
Subject: Re: [R] Durbin-Watson test in packages "car" and
"lmtest"
> Hi,
> P-values in Durbin-Watson test obtained through the use of
> functions available in packages "lmtest" and "car" are
different. The
> difference is quite significant. function "dwtest" in
"lmtest" is much
> faster than "burbinwatson" in "car". Actually, you can
take a nap while
> the latter trying to calculated Durbin-Watson test. My question is which
> p-value is better?
The answer is essencially given in ?durbin.watson and ?dwtest. The latter
states that
The p value is computed
using a Fortran version of the Applied Statistics Algorithm AS 153
by Farebrother (1980, 1984). This algorithm is called "pan" or
"gradsol". For large sample sizes the algorithm might fail to
compute the p value; in that case a warning is printed and an
approximate p value will be given; this p value is computed using
a normal approximation with mean and variance of the Durbin-Watson
test statistic.
while ?durbin.watson says
simulate: if `TRUE' p-values will be estimated by bootstrapping.
What is a "quite significant" difference for p-values?
Looking at the example from ?durbin.watson gives:
R> durbin.watson(lm(fconvict ~ tfr + partic + degrees + mconvict,
data=Hartnagel))
lag Autocorrelation D-W Statistic p-value
1 0.688345 0.6168636 0
R> dwtest(fconvict ~ tfr + partic + degrees + mconvict, data=Hartnagel)
Durbin-Watson test
data: fconvict ~ tfr + partic + degrees + mconvict
DW = 0.6169, p-value = 6.96e-09
which is fairly close, so you might give us more details (that is: a
working example) to see what the "difference" is (that is: a bug in
either function or a difference due to simulation error / bad
approximation ...).
Torsten
>
> Thank you,
> Ahmad Abu Hammour
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
https://stat.ethz.ch/pipermail/r-help/attachments/20020419/85bd6b79/attachment.html
Lu Chi-Hsien Joseph
2002-Apr-20 02:17 UTC
[R] Durbin-Watson test in packages "car" and "lmtest"
The difference that I found in applying Durbin-Watson test in "car" and "lmtest" is the latter only works for testing NH: rho = 0 versus AH: rho > 0. C. Joseph Lu Department of Statistics National Cheng-Kung University Tainan, Taiwan -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Dear Ahmad, I think that the discrepancy has been cleared up, since durbin.watson, as has been pointed out, calculates two-sided tests. Some other relevant points: (1) durbin.watson will calculate dw statistics and associated p-values for lags greater than 1. (2) I prefer the approach currently in dwtest, but when durbin.watson was written, dwtest in lmtest didn't compute p-values. (3) I also agree that it would be useful to allow one-sided alternatives in durbin.watson and will likely add this if I don't remove the function from the car package. I apologize for my slow response to your original posting and the subsequent exchange -- I was out of town. John At 05:05 PM 4/19/2002 -0400, Ahmad Abu Hammour wrote:>Hi Torsten, >Here is an example in which P-values have a significant difference. The >data is a sub sample of a larger one. >Basically the model I am using is as follows: >y_{t+2}= y_{t+1} + y_t + I(x_{t+1}-z_{t+1}-w_{t+1}) +e_{t+2} > > > ydata > y z w x > [1,] 0.0101705342 -5.363636567 0.80677087 -2.425968 > [2,] -0.0040963887 -4.930336567 -0.01515276 -1.852668 > [3,] 0.0007996203 -2.795136567 0.28412361 -0.849268 > [4,] 0.0231375722 -2.523036567 -0.33650002 -0.315968 > [5,] 0.0576658597 -2.925936567 -0.68112365 0.060732 > [6,] 0.0930807057 -2.283836567 -1.59374728 0.214032 > [7,] 0.0904861271 -1.908436567 -3.15057091 0.347332 > [8,] 0.0815802452 -3.063936567 -3.60669454 -0.175968 > [9,] 0.0463679856 -3.947236567 -2.89361816 -1.012668 >[10,] 0.0283314000 -3.528136567 -1.68974179 -1.522668 >[11,] 0.0158693958 -2.890536567 -1.03456542 -1.629268 >[12,] -0.0177988767 -2.165736567 -0.23008905 -1.532668 >[13,] -0.0263232785 -2.400436567 -0.29991268 -1.369268 >[14,] -0.0377695478 -3.243136567 -0.31953631 -0.539268 >[15,] -0.0520095622 -3.241336567 0.22174006 0.120732 >[16,] -0.0773138815 -2.765236567 -0.22328356 1.467332 >[17,] -0.0804307590 -2.833236567 0.09209281 2.107332 >[18,] -0.0446189151 -3.710136567 0.57246918 2.057332 >[19,] -0.0439761920 -4.232636567 0.11764555 2.687332 >[20,] -0.0508314286 -3.445136567 0.91252192 4.874032 >[21,] -0.0373526376 -2.123836567 1.03269829 5.340732 >[22,] -0.0363366045 -4.549336567 1.16137466 3.614032 >[23,] -0.0119795294 -4.660636567 2.10495103 1.744032 >[24,] -0.0271166745 0.940863433 2.73172741 5.450732 >[25,] -0.0319279899 2.972463433 3.93230378 7.954032 >[26,] -0.0285519284 2.601663433 4.25698015 9.447332 >[27,] -0.0362890374 3.900363433 4.49975652 11.397332 >[28,] -0.0193261619 2.440563433 4.24003289 7.057332 >[29,] -0.0211604332 4.133063433 3.62460926 5.907332 >[30,] -0.0393599621 4.764863433 3.78798563 6.740732 >[31,] -0.0368496844 2.967363433 3.09016201 5.137332 >[32,] -0.0111656400 2.374163433 2.32173838 1.820732 >[33,] -0.0049543837 2.858663433 0.59471475 0.570732 >[34,] -0.0085112660 3.900063433 -0.96020888 0.424032 >[35,] 0.0002988086 3.956163433 -1.40353251 0.500732 >[36,] -0.0021387277 2.276463433 -1.99135614 0.720732 >[37,] -0.0143497002 2.857863433 -1.40157977 1.270732 >[38,] -0.0459512047 3.589163433 -1.78060340 2.574032 >[39,] -0.0620556482 4.248163433 -2.43422702 3.537332 >[40,] -0.0665723041 3.225363433 -2.45305065 1.830732 >[41,] -0.0990262190 2.638163433 -2.34577428 1.634032 >[42,] -0.0993943457 2.160963433 -2.00509791 0.780732 >[43,] -0.0896954664 1.651763433 -1.79992154 0.164032 >[44,] -0.1047509912 2.106963433 -1.59274517 0.117332 >[45,] -0.1188326548 3.230463433 -1.40576880 2.007332 >[46,] -0.1012159404 2.484063433 -1.60469242 -0.209268 >[47,] -0.0975617833 2.185163433 -1.23841605 -0.442668 >[48,] -0.0859923186 1.244563433 -0.99983968 -0.495968 >[49,] -0.0467949427 -0.185636567 -1.19776331 -1.649268 >[50,] -0.0351160760 -0.510836567 -0.50778694 -0.569268 >[51,] -0.0233943108 -0.356136567 -0.53781057 0.347332 >[52,] -0.0080094290 0.005063433 -0.71733420 -0.569268 >[53,] 0.0312603960 -0.149336567 -0.69205783 -0.349268 >[54,] 0.0614944090 0.181063433 -0.76188145 0.237332 >[55,] 0.0775709808 0.795863433 -0.58080508 1.110732 >[56,] 0.0952235601 1.131763433 -0.38712871 1.907332 >[57,] 0.1069545973 1.509763433 0.06534766 2.887332 >[58,] 0.1136253179 1.785763433 0.64482403 3.450732 >[59,] 0.1275201332 1.348463433 1.06650040 3.410732 >[60,] 0.1387864450 0.581263433 1.05767677 3.444032 >[61,] 0.1244438565 1.253563433 1.35555315 4.164032 >[62,] 0.1344048716 0.334263433 0.69162952 4.844032 >[63,] 0.1487305153 -0.634036567 0.37670589 4.010732 >[64,] 0.1420063384 -0.159036567 1.22928226 3.190732 >[65,] 0.1469999823 -0.707536567 1.67200000 1.200732 > > library(car) >Attaching package `car': > > The following object(s) are masked from package:base : > dfbetas rstudent > > library(lmtest) > > > lmy=lm(y[-c(1,2)]~y[-c(1,65)]+y[-c(64,65)]+I(x-z-w)[-c(64,65)]-1,data=as.ts(ydata)) > > dwtest(lmy) > Durbin-Watson test >data: lmy >DW = 2.0077, p-value = 0.4404 > > durbin.watson(lmy) > lag Autocorrelation D-W Statistic p-value > 1 -0.01370151 2.007701 0.92 > > >So, which P-value should I adopt 0.44 or 0.92? >Thank you for your help. >Ahmad Abu Hammour > >----- Original Message ----- >From: Torsten Hothorn >Sent: Friday, April 19, 2002 4:51 AM >To: Ahmad Abu Hammour >Cc: R-help at stat.math.ethz.ch >Subject: Re: [R] Durbin-Watson test in packages "car" and "lmtest" > > > Hi, > > P-values in Durbin-Watson test obtained through the use of > > functions available in packages "lmtest" and "car" are different. The > > difference is quite significant. function "dwtest" in "lmtest" is much > > faster than "burbinwatson" in "car". Actually, you can take a nap while > > the latter trying to calculated Durbin-Watson test. My question is which > > p-value is better? > >The answer is essencially given in ?durbin.watson and ?dwtest. The latter >states that > >The p value is computed > using a Fortran version of the Applied Statistics Algorithm AS 153 > by Farebrother (1980, 1984). This algorithm is called "pan" or > "gradsol". For large sample sizes the algorithm might fail to > compute the p value; in that case a warning is printed and an > approximate p value will be given; this p value is computed using > a normal approximation with mean and variance of the Durbin-Watson > test statistic. >while ?durbin.watson says > > simulate: if `TRUE' p-values will be estimated by bootstrapping. > >What is a "quite significant" difference for p-values? > >Looking at the example from ?durbin.watson gives: > >R> durbin.watson(lm(fconvict ~ tfr + partic + degrees + mconvict, >data=Hartnagel)) >lag Autocorrelation D-W Statistic p-value > 1 0.688345 0.6168636 0 >R> dwtest(fconvict ~ tfr + partic + degrees + mconvict, data=Hartnagel) > > Durbin-Watson test > >data: fconvict ~ tfr + partic + degrees + mconvict >DW = 0.6169, p-value = 6.96e-09 > >which is fairly close, so you might give us more details (that is: a >working example) to see what the "difference" is (that is: a bug in >either function or a difference due to simulation error / bad >approximation ...). > >Torsten > > > > > Thank you, > > Ahmad Abu Hammour > >----------------------------------------------------- John Fox Department of Sociology McMaster University Hamilton, Ontario, Canada L8S 4M4 email: jfox at mcmaster.ca phone: 905-525-9140x23604 web: www.socsci.mcmaster.ca/jfox ----------------------------------------------------- -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._