thr3ads.net - R help - [R] anova.rq {quantreg) - Why do different level of nesting changes the P values?! [Oct 2011]

If this information is useful, please help other people find it:
Share via:

Tal Galili

2011-Oct-06 08:05 UTC

[R] anova.rq {quantreg) - Why do different level of nesting changes the P values?!

Hello dear R help members.

I am trying to understand the anova.rq, and I am finding something which I
can not explain (is it a bug?!):

The example is for when we have 3 nested models.  I run the anova once on
the two models, and again on the three models.  I expect that the p.value
for the comparison of model 1 and model 2 would remain the same, whether or
not I add a third model to be compared with.
However, the P values change, and I do not understand why.

Here is an example code (following with it's input):

data(barro)
fit0 <- rq(y.net ~  lgdp2 + fse2 , data = barro)
fit1 <- rq(y.net ~  lgdp2 + fse2 + gedy2 , data = barro)
fit2 <- rq(y.net ~  lgdp2 + fse2 + gedy2 + Iy2 , data = barro)
anova(fit0,fit1,fit2, R = 1000)
anova(fit0,fit1, R = 1000)


Output:
> data(barro)
> fit0 <- rq(y.net ~  lgdp2 + fse2 , data = barro)
> fit1 <- rq(y.net ~  lgdp2 + fse2 + gedy2 , data = barro)
> fit2 <- rq(y.net ~  lgdp2 + fse2 + gedy2 + Iy2 , data = barro)
> anova(fit0,fit1,fit2, R = 1000)Quantile Regression Analysis of Deviance Table

Model 1: y.net ~ lgdp2 + fse2 + gedy2 + Iy2
Model 2: y.net ~ lgdp2 + fse2 + gedy2
Model 3: y.net ~ lgdp2 + fse2
  Df Resid Df F value    Pr(>F)
1  1      156  29.494 2.110e-07 ***
2  2      156  18.194 7.901e-08 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’
1> anova(fit0,fit1, R = 1000)Quantile Regression Analysis of Deviance Table

Model 1: y.net ~ lgdp2 + fse2 + gedy2
Model 2: y.net ~ lgdp2 + fse2
  Df Resid Df F value  Pr(>F)
1  1      157  3.9532 0.04852 *
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’
1> sessionInfo()R version 2.13.1 (2011-07-08)
Platform: i386-pc-mingw32/i386 (32-bit)

locale:
[1] LC_COLLATE=Hebrew_Israel.1255  LC_CTYPE=Hebrew_Israel.1255
[3] LC_MONETARY=Hebrew_Israel.1255 LC_NUMERIC=C
[5] LC_TIME=Hebrew_Israel.1255

attached base packages:
[1] splines   stats     graphics  grDevices utils     datasets  methods
base

other attached packages:
 [1] rms_3.3-1                    Hmisc_3.8-3
 survival_2.36-9
 [4] colorspace_1.1-0             quantreg_4.71                SparseM_0.89

 [7] PerformanceAnalytics_1.0.3.2 xts_0.8-2                    zoo_1.7-4

[10] reporttools_1.0.6            xtable_1.5-6

loaded via a namespace (and not attached):
[1] cluster_1.14.0  grid_2.13.1     lattice_0.19-33 tools_2.13.1






----------------Contact
Details:-------------------------------------------------------
Contact me: Tal.Galili@gmail.com |  972-52-7275845
Read me: www.talgalili.com (Hebrew) | www.biostatistics.co.il (Hebrew) |
www.r-statistics.com (English)
----------------------------------------------------------------------------------------------

	[[alternative HTML version deleted]]

Eik Vettorazzi

2011-Oct-06 11:21 UTC

head link

[R] anova.rq {quantreg) - Why do different level of nesting changes the P values?!

Hi Tal,
you are comparing different things. The "Details"-section of ?anova.qr
states "test the hypothesis that smaller models are adequate relative to
the largest specified model". So in your first anova you compare fit2
with fit1 and fit0, in your second attempt fit1 with fit0, so you have
different "base"-models to compare with, and consequently different
p-values.

hth.

Am 06.10.2011 10:05, schrieb Tal Galili:> Hello dear R help members.
> 
> I am trying to understand the anova.rq, and I am finding something which I
> can not explain (is it a bug?!):
> 
> The example is for when we have 3 nested models.  I run the anova once on
> the two models, and again on the three models.  I expect that the p.value
> for the comparison of model 1 and model 2 would remain the same, whether or
> not I add a third model to be compared with.
> However, the P values change, and I do not understand why.
> 
> Here is an example code (following with it's input):
> 
> data(barro)
> fit0 <- rq(y.net ~  lgdp2 + fse2 , data = barro)
> fit1 <- rq(y.net ~  lgdp2 + fse2 + gedy2 , data = barro)
> fit2 <- rq(y.net ~  lgdp2 + fse2 + gedy2 + Iy2 , data = barro)
> anova(fit0,fit1,fit2, R = 1000)
> anova(fit0,fit1, R = 1000)
> 
> 
> Output:
> 
>> data(barro)
>> fit0 <- rq(y.net ~  lgdp2 + fse2 , data = barro)
>> fit1 <- rq(y.net ~  lgdp2 + fse2 + gedy2 , data = barro)
>> fit2 <- rq(y.net ~  lgdp2 + fse2 + gedy2 + Iy2 , data = barro)
>> anova(fit0,fit1,fit2, R = 1000)
> Quantile Regression Analysis of Deviance Table
> 
> Model 1: y.net ~ lgdp2 + fse2 + gedy2 + Iy2
> Model 2: y.net ~ lgdp2 + fse2 + gedy2
> Model 3: y.net ~ lgdp2 + fse2
>   Df Resid Df F value    Pr(>F)
> 1  1      156  29.494 2.110e-07 ***
> 2  2      156  18.194 7.901e-08 ***
> ---
> Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
>> anova(fit0,fit1, R = 1000)
> Quantile Regression Analysis of Deviance Table
> 
> Model 1: y.net ~ lgdp2 + fse2 + gedy2
> Model 2: y.net ~ lgdp2 + fse2
>   Df Resid Df F value  Pr(>F)
> 1  1      157  3.9532 0.04852 *
> ---
> Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
>> sessionInfo()
> R version 2.13.1 (2011-07-08)
> Platform: i386-pc-mingw32/i386 (32-bit)
> 
> locale:
> [1] LC_COLLATE=Hebrew_Israel.1255  LC_CTYPE=Hebrew_Israel.1255
> [3] LC_MONETARY=Hebrew_Israel.1255 LC_NUMERIC=C
> [5] LC_TIME=Hebrew_Israel.1255
> 
> attached base packages:
> [1] splines   stats     graphics  grDevices utils     datasets  methods
> base
> 
> other attached packages:
>  [1] rms_3.3-1                    Hmisc_3.8-3
>  survival_2.36-9
>  [4] colorspace_1.1-0             quantreg_4.71                SparseM_0.89
> 
>  [7] PerformanceAnalytics_1.0.3.2 xts_0.8-2                    zoo_1.7-4
> 
> [10] reporttools_1.0.6            xtable_1.5-6
> 
> loaded via a namespace (and not attached):
> [1] cluster_1.14.0  grid_2.13.1     lattice_0.19-33 tools_2.13.1
> 
> 
> 
> 
> 
> 
> ----------------Contact
> Details:-------------------------------------------------------
> Contact me: Tal.Galili at gmail.com |  972-52-7275845
> Read me: www.talgalili.com (Hebrew) | www.biostatistics.co.il (Hebrew) |
> www.r-statistics.com (English)
>
----------------------------------------------------------------------------------------------
> 
> 	[[alternative HTML version deleted]]
> 
> 
> 
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

-- 
Eik Vettorazzi

Department of Medical Biometry and Epidemiology
University Medical Center Hamburg-Eppendorf

Martinistr. 52
20246 Hamburg

T ++49/40/7410-58243
F ++49/40/7410-57790

--
Pflichtangaben gem?? Gesetz ?ber elektronische Handelsregister und
Genossenschaftsregister sowie das Unternehmensregister (EHUG):

Universit?tsklinikum Hamburg-Eppendorf; K?rperschaft des ?ffentlichen Rechts;
Gerichtsstand: Hamburg

Vorstandsmitglieder: Prof. Dr. Guido Sauter (Vertreter des Vorsitzenden), Dr.
Alexander Kirstein, Joachim Pr?l?, Prof. Dr. Dr. Uwe Koch-Gromus

Maybe Matching Threads

Search for more maybe matching threads

R help - Oct 2011 - anova.rq {quantreg) - Why do different level of nesting changes the P values?!

[R] anova.rq {quantreg) - Why do different level of nesting changes the P values?!

[R] anova.rq {quantreg) - Why do different level of nesting changes the P values?!

Maybe Matching Threads