Skip to content

[R-meta] A question about meta-regression Analysis

7 messages · Wolfgang Viechtbauer, Michael Dewey, Hamed Kianmehr

#
I have some questions regarding the meta-regression analysis. I have the
database on the proportion (Below bubble plot). My advisor asked me to do
the polynomial regression. I have two questions:

1- Is it any polynomial regression pattern in this data?
2- I wrote below code but I couldn't get the line, Could you help me?
----------------
res<-rma(yi=y, vi=s2, data=data, method="DL",mods = ~ poly(year, degree=2))
years <- 1989:2016
preds <- predict(res, transf = transf.ilogit, newmods =cbind(years,years^2))
wi <- 0.5/sqrt(data$s2)
wi <- 0.5 + 3.0 * (wi - min(wi))/(max(wi) - min(wi))
plot(data$year, transf.ilogit(y), cex=wi,xlab="Year",ylab="Patient
Expectation (%)")
lines(years, preds$pred)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://stat.ethz.ch/pipermail/r-sig-meta-analysis/attachments/20180713/751d364d/attachment-0001.html>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rplot.jpeg
Type: image/jpeg
Size: 35122 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-meta-analysis/attachments/20180713/751d364d/attachment-0001.jpeg>
#
Dear Hamed,

Please post in plain text format. I cannot see any plot. Either attach it as a separate image file or post it somewhere else and provide a link.

As for 2 - If you use poly(), try:

preds <- predict(res, transf = transf.ilogit, newmods = unname(poly(years, degree=2)))

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Hamed Kianmehr
Sent: Friday, 13 July, 2018 16:03
To: r-sig-meta-analysis at r-project.org
Subject: [R-meta] A question about meta-regression Analysis

I have some questions regarding the meta-regression analysis. I have the database on the proportion (Below bubble plot). My advisor?asked me to do the polynomial?regression. I have two questions:

1- Is it any polynomial regression pattern in this data?
2- I wrote below code but I couldn't get the line, Could you help me?
----------------
res<-rma(yi=y, vi=s2, data=data, method="DL",mods = ~ poly(year, degree=2))
years <- 1989:2016
preds <- predict(res, transf = transf.ilogit, newmods =cbind(years,years^2))
wi <- 0.5/sqrt(data$s2)
wi <- 0.5 + 3.0 * (wi - min(wi))/(max(wi) - min(wi))
plot(data$year, transf.ilogit(y), cex=wi,xlab="Year",ylab="Patient Expectation (%)")
lines(years, preds$pred)
#
Thank you for your response.
Please see the attached file regarding the plot.

On Fri, Jul 13, 2018 at 12:35 PM, Viechtbauer, Wolfgang (SP) <
wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:

            

  
    
#
Thanks. Well, hard to say if these data are suggestive of a quadratic trend based on visual inspection.

The test of the coefficient for 'poly(year, 2)2' tells you whether the quadratic model fits significantly better than just the linear model.

Best,
Wolfgang

-----Original Message-----
From: Hamed Kianmehr [mailto:hkianme1 at binghamton.edu] 
Sent: Friday, 13 July, 2018 18:39
To: Viechtbauer, Wolfgang (SP)
Cc: r-sig-meta-analysis at r-project.org
Subject: Re: [R-meta] A question about meta-regression Analysis

ATTACHMENT(S) REMOVED: Plot.jpg 

Thank you for your response.?
Please see the attached?file regarding the plot.?
On Fri, Jul 13, 2018 at 12:35 PM, Viechtbauer, Wolfgang (SP) <wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:
Dear Hamed,

Please post in plain text format. I cannot see any plot. Either attach it as a separate image file or post it somewhere else and provide a link.

As for 2 - If you use poly(), try:

preds <- predict(res, transf = transf.ilogit, newmods = unname(poly(years, degree=2)))

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Hamed Kianmehr
Sent: Friday, 13 July, 2018 16:03
To: r-sig-meta-analysis at r-project.org
Subject: [R-meta] A question about meta-regression Analysis

I have some questions regarding the meta-regression analysis. I have the database on the proportion (Below bubble plot). My advisor?asked me to do the polynomial?regression. I have two questions:

1- Is it any polynomial regression pattern in this data?
2- I wrote below code but I couldn't get the line, Could you help me?
----------------
res<-rma(yi=y, vi=s2, data=data, method="DL",mods = ~ poly(year, degree=2))
years <- 1989:2016
preds <- predict(res, transf = transf.ilogit, newmods =cbind(years,years^2))
wi <- 0.5/sqrt(data$s2)
wi <- 0.5 + 3.0 * (wi - min(wi))/(max(wi) - min(wi))
plot(data$year, transf.ilogit(y), cex=wi,xlab="Year",ylab="Patient Expectation (%)")
lines(years, preds$pred)

--
Best Regards,
Hamed Kianmehr, Ph.D. Candidate
Department of Systems Science and Industrial Engineering
The State University of New York at Binghamton
#
Thank you for your prompt response.

For the test, do you mean the value of coefficient(s) (attached file)?  if
yes, how can I find it better than rest of them?

On Fri, Jul 13, 2018 at 1:02 PM, Viechtbauer, Wolfgang (SP) <
wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:

            

  
    
1 day later
#
Dear Hamed

If the line is supposed to show a polynomial of degree two then to my 
eyes it does not. It looks linear. If you think there may be a quadratic 
term why do you not fit two models, a linear on and a quadratic one, and 
just compare them with anova()?

Most of the points on your plot lie on one side of the line (above) 
which does not look quite right.

Michael
On 13/07/2018 17:38, Hamed Kianmehr wrote:

  
    
  
#
Thank you very much, Michael. So, I'll look into it with anova.

What does it mean " Most of the points on your plot lie on one side of the
line (above) which does not look quite right."? is there any way to fix it?





On Sun, Jul 15, 2018 at 12:07 PM, Michael Dewey <lists at dewey.myzen.co.uk>
wrote: