Skip to content

How to make a nomogam and Calibration plot

6 messages · sytangping, Frank E Harrell Jr, zhu yao

#
Dear R users,

I am a new R user and something stops me when I try to write a academic
article. I want to make a nomogram to predict the risk of prostate cancer
(PCa) using several factors which have been selected from the Logistic
regression run under the SPSS. Always, a calibration plot is needed to
validate the prediction accuracy of the nomogram.
However, I tried many times and read a lot of posts with respect to this
topic but I still couldn't figure out how to draw the nomogram and the
calibration plot. My dataset and questions in detail are shown in two
attached files. It will be very grateful if someone can save his/her time to
help for my questions.

Warmest regards!

Ping Tang   http://r.789695.n4.nabble.com/file/n3710068/Dataset.xls
Dataset.xls  http://r.789695.n4.nabble.com/file/n3710068/R_help.doc
R_help.doc 

--
View this message in context: http://r.789695.n4.nabble.com/How-to-make-a-nomogam-and-Calibration-plot-tp3710068p3710068.html
Sent from the R help mailing list archive at Nabble.com.
#
Kindly do not attach questions in a separate document.

Install and read the documentation for the R rms package, and see handouts
at http://biostat.mc.vanderbilt.edu/rms

Frank
sytangping wrote:
-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/How-to-make-a-nomogam-and-Calibration-plot-tp3710068p3710126.html
Sent from the R help mailing list archive at Nabble.com.
1 day later
#
Dear Harrell,

Many thanks for your quick response!
However, after try and try, I still have difficulty to solve my questions.
I post my questions again. I hope someone can help me run the data and draw
the nomogram and calibration plot for me. I know that is not good but indeed
I have no way to go. The problems almost drove me mad!

Best regards!

Ping Tang

Dear R users,

I am a new R user and something stops me when I try to write a academic
article. I want to make a nomogram to predict the risk of prostate cancer
(PCa) using several factors which have been selected from the Logistic
regression run under the SPSS. Always, a calibration plot is needed to
validate the prediction accuracy of the nomogram.
However, I tried many times and read a lot of posts with respect to this
topic but I still couldn't figure out how to draw the nomogram and the
calibration plot. Attached file is the dataset for the research. It will be
very grateful if someone can save his/her time to help for my questions.

Warmest regards!

Logistic Regression

Classification Tablea,b
	Observed	Predicted
		Pca-YN	Percentage Correct
		0	1	
Step 0	Pca-YN	0	295	0	100.0
		1	218	0	.0
	Overall Percentage			57.5


Variables in the Equation
		B	S.E.	Wald	df	Sig.	Exp(B)	95.0% C.I.for EXP(B)
								Lower	Upper
Step 1a	Age	.031	.015	4.491	1	.034	1.032	1.002	1.062
	DRE	1.173	.266	19.492	1	.000	3.233	1.920	5.443
	LogPV	-2.857	.509	31.532	1	.000	.057	.021	.156
	LogPSA	2.316	.246	88.416	1	.000	10.132	6.253	16.419
	Constant	-1.024	1.273	.648	1	.421	.359		
				

The equation:

Probability = 

 e-1.024+0.31age+1.173DRE+-2.857LogPV+2.316LogPSA
1+e-1.024+0.31age+1.173DRE+-2.857LogPV+2.316LogPSA

My questions are,

1.How to draw a nomogram (similar to the below figure 1) to predict the
probability of cancer using R?

2. How to make the Calibration plot (similar to the below figure 2) which
used to validate the prediction accuracy of the nomogram using R? And how to
calculate the concordance index (C-index) ?


http://r.789695.n4.nabble.com/file/n3714477/untitled.jpg 
http://r.789695.n4.nabble.com/file/n3714477/%E9%99%84%E4%BB%B62.jpg 
http://r.789695.n4.nabble.com/file/n3714477/Dataset.xls Dataset.xls 

--
View this message in context: http://r.789695.n4.nabble.com/How-to-make-a-nomogam-and-Calibration-plot-tp3710068p3714477.html
Sent from the R help mailing list archive at Nabble.com.
#
The nomogram you included was produced by the Design package, the precursor
to the rms package.  You will have to take the time to intensively read the
rms package documentation.  Note that how you developed the model (e.g.,
allowing for non-linearity in log PSA, not using stepwise regression which
invalidates the results, making sure all clinically relevant predictors are
in the model, ...) is the most important step.  The process you are going
through generally requires an M.S. in biostatistics.
Frank
sytangping wrote:
-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/How-to-make-a-nomogam-and-Calibration-plot-tp3710068p3715336.html
Sent from the R help mailing list archive at Nabble.com.
#
I want to add a comment related to the calibration plot that was presented in
a previous post (which probably cannot be done optimally in SPSS).  The plot
lacks sufficient resolution in the x-axis values that are calibrated.  It is
far better to use loess to estimate a smooth nonparametric calibration
curve, with no (arbitrary) binning of x.  And it is not adequate to validate
only the 3-4 points that were plotted.
Frank
yz wrote:
-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/How-to-make-a-nomogam-and-Calibration-plot-tp3710068p3717519.html
Sent from the R help mailing list archive at Nabble.com.