thr3ads.net - R help - [R] I am struggling with contrasts [Mar 2020]

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

peter dalgaard

2020-Mar-10 09:07 UTC

[R] I am struggling with contrasts

Yes. Contrasts, by definition, represents between-group differences, so cannot
yield individual group levels. The closest you get is that the _intercept_ is
the level of the base group in treatment contrasts, and relevel() will allow you
to change the base level.

-pd
> On 10 Mar 2020, at 04:06 , Berwin A Turlach <Berwin.Turlach at
gmail.com> wrote:
> 
> G'day John,
> 
> On Tue, 10 Mar 2020 01:42:46 +0000
> "Sorkin, John" <jsorkin at som.umaryland.edu> wrote:
> 
>> I am running a Poisson regression with a single outcome variable,
>> HGE, and a single independent variable, a factor, Group which can be
>> one of two values, Group1, or Group2. I am trying to define contrasts
>> that will give me the values of my outcome variable (HGE) when
>> group=Group1 and when group=Group2. 
> 
> Not sure what you mean, but I am suspecting you are after this output:
> 
> R> fit0 <- glm(HGE ~ Group -
1,family=poisson,data=dataForR,offset=logFU)
> R> summary(fit0)
> 
> Cheers,
> 
> 	Berwin
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

Sorkin, John

2020-Mar-10 10:50 UTC

head link

[R] I am struggling with contrasts

I have not clearly stated my question. I would like to obtain the point estimate
and SE (or point estimate and 95% CI) of a linear combination of the the
independent variables included in my regression model. In a simple model having
a single categorical variable that has two levels (Group1 and Group2) obtaining
the estimate and its SE (or the estimate and a 95% CI) requires knowing the
betas produced by the model, the SEs of the betas (which are easily obtained)
along with the variance covariance of the estimates. I assume that the variance
covariance matrix can be obtained but working the the matrix is a real pain. I
am looking for a SIMPLE way to get the point estimate and its SE without having
to slog though getting all the estimates, their SEs manually adding them
together and including the covariances.
For example if my model is

rate = group and group has the value 1, I want:

beta rate = beta intercept + beta group
variance rate = variance intercept + variance group + 2*covariance
(intercept,group)

I suspect I can do this calculation manually, but I would really like to find a
way that R will do the computation for me.

My regression model is:
fit1 <- glm(HGE ~ Group,family=quasipoisson(link="log"),
data=dataForR,offset=logFU)

In SAS this can be accomplished using estimate statements; I suspect that is an
R analogue of the SAS estimate statement, but I don't know that the analogue
is .

Thank you,

John

Particular thanks are due to Peter Dalgaard, Berwin Turlach, and Mark Leeds who
responded to my original, not well formulated posting.

Thank you,

John

John David Sorkin M.D., Ph.D.
Professor of Medicine
Chief, Biostatistics and Informatics
University of Maryland School of Medicine Division of Gerontology and Geriatric
Medicine
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524
(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)

________________________________
From: peter dalgaard <pdalgd at gmail.com>
Sent: Tuesday, March 10, 2020 5:07 AM
To: Berwin A Turlach <Berwin.Turlach at gmail.com>
Cc: Sorkin, John <jsorkin at som.umaryland.edu>; r-help at r-project.org
(r-help at r-project.org) <r-help at r-project.org>
Subject: Re: [R] I am struggling with contrasts

Yes. Contrasts, by definition, represents between-group differences, so cannot
yield individual group levels. The closest you get is that the _intercept_ is
the level of the base group in treatment contrasts, and relevel() will allow you
to change the base level.

-pd
> On 10 Mar 2020, at 04:06 , Berwin A Turlach <Berwin.Turlach at
gmail.com> wrote:
>
> G'day John,
>
> On Tue, 10 Mar 2020 01:42:46 +0000
> "Sorkin, John" <jsorkin at som.umaryland.edu> wrote:
>
>> I am running a Poisson regression with a single outcome variable,
>> HGE, and a single independent variable, a factor, Group which can be
>> one of two values, Group1, or Group2. I am trying to define contrasts
>> that will give me the values of my outcome variable (HGE) when
>> group=Group1 and when group=Group2.
>
> Not sure what you mean, but I am suspecting you are after this output:
>
> R> fit0 <- glm(HGE ~ Group -
1,family=poisson,data=dataForR,offset=logFU)
> R> summary(fit0)
>
> Cheers,
>
>        Berwin
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>
https://nam03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fstat.ethz.ch%2Fmailman%2Flistinfo%2Fr-help&amp;data=02%7C01%7C%7C4b9ea7f815e0494abf0908d7c4d2770c%7C717009a620de461a88940312a395cac9%7C0%7C0%7C637194280543093844&amp;sdata=QFyKBzxZfkXAjASSUYsSnMJ1CiZIjXDK6JEgKHrV5vg%3D&amp;reserved=0
> PLEASE do read the posting guide
https://nam03.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.R-project.org%2Fposting-guide.html&amp;data=02%7C01%7C%7C4b9ea7f815e0494abf0908d7c4d2770c%7C717009a620de461a88940312a395cac9%7C0%7C0%7C637194280543093844&amp;sdata=pf%2BQVOSaoGq5HwJ1QjMawCArFC1yjpG9yL%2Fs5dv5wJw%3D&amp;reserved=0
> and provide commented, minimal, self-contained, reproducible code.
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

	[[alternative HTML version deleted]]

Viechtbauer, Wolfgang (SP)

2020-Mar-10 11:07 UTC

head link

[R] I am struggling with contrasts

Dear John,

The linearHypothesis() function from the 'car' package does this.
>From the help file: "The value of the linear hypothesis and its
covariance matrix are returned respectively as "value" and
"vcov" attributes of the object (but not printed)." For  a single
linear combination, vcov will be a single value and its square-root the SE.
Best,
Wolfgang

-----Original Message-----
From: R-help [mailto:r-help-bounces at r-project.org] On Behalf Of Sorkin, John
Sent: Tuesday, 10 March, 2020 11:51
To: peter dalgaard; Berwin A Turlach
Cc: r-help at r-project.org (r-help at r-project.org)
Subject: Re: [R] I am struggling with contrasts

I have not clearly stated my question. I would like to obtain the point estimate
and SE (or point estimate and 95% CI) of a linear combination of the the
independent variables included in my regression model. In a simple model having
a single categorical variable that has two levels (Group1 and Group2) obtaining
the estimate and its SE (or the estimate and a 95% CI) requires knowing the
betas produced by the model, the SEs of the betas (which are easily obtained)
along with the variance covariance of the estimates. I assume that the variance
covariance matrix can be obtained but working the the matrix is a real pain. I
am looking for a SIMPLE way to get the point estimate and its SE without having
to slog though getting all the estimates, their SEs manually adding them
together and including the covariances.
For example if my model is

rate = group and group has the value 1, I want:

beta rate = beta intercept + beta group
variance rate = variance intercept + variance group + 2*covariance
(intercept,group)

I suspect I can do this calculation manually, but I would really like to find a
way that R will do the computation for me.

My regression model is:
fit1 <- glm(HGE ~ Group,family=quasipoisson(link="log"),
data=dataForR,offset=logFU)

In SAS this can be accomplished using estimate statements; I suspect that is an
R analogue of the SAS estimate statement, but I don't know that the analogue
is .

Thank you,

John

Particular thanks are due to Peter Dalgaard, Berwin Turlach, and Mark Leeds who
responded to my original, not well formulated posting.

Thank you,

John

John David Sorkin M.D., Ph.D.
Professor of Medicine
Chief, Biostatistics and Informatics
University of Maryland School of Medicine Division of Gerontology and Geriatric
Medicine
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524
(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)

R help - Mar 2020 - I am struggling with contrasts

[R] I am struggling with contrasts

[R] I am struggling with contrasts

[R] I am struggling with contrasts