Skip to main content

Modeling gender and age adjusted incidence rates

National Health Institute (NHI) provides a tool box for calculation of cancer incidence and percentage change. Their algorithm for Jointpoint Trend Analysis is well-documented but does not provide the best tool at hand for most problems. The normal approximation is not the most optimal choice for situations with a low incidence rate in which I would recommend to apply modern logistic regression algorithms which are far more versatile.

 

The difference between careful parametrization in a binomial regression model and the plug-and-play functionality of the NHI suite becomes obvious in an example in which we look at cancers in children. Data source: NORDCAN

Logistic regression models. Joint point model (left) using stepwise linear gender specific regression models and polynomial models (right) using gender specific polynomial regression models.
 


Graphs with gender specific 95% prediction limits
 
 
 
 


R-script Data Extraction
SAS program

Joint Point Model based on software from NIH
The estimation procedure does not allow zero-counts, which introduces bias.


Furthermore, errors are approximate normal distributed.


The logistic regression model predicts a total of 190 cancer cases during the period 1979-2014, whereas the Jointpoint trend program from NHI predicts 158 cases of cancer when adjusting for calendar year. Binomial model estimates a total combined incidence rate of 0.57 (per 100,000) corrected for calendar year, whereas the Jointpoint trend analysis program yields an incidence rate of 0.47 (per 100,000). We observe a total of 33,679,014 person years.
We have used actual connective tissue cancer incidence counts for Danes age 0 to 24 from the NORDCAN register of gender specific incidence rates with a total of 189 cases in the period 1979-2014.


Comments

Popular posts from this blog

HackRF on Windows 8

This technical note is based on an extract from thread. I have made several changes and added recommendations. I have experienced lot of latency using GnuRadio and HackRF on Pentoo Linux, so I wanted to try out GnuRadio on Windows.



HackRF One is a transceiver, so besides SDR capabilities, it can also transmit signals, inkluding sweeping a given range, uniform and Gaussian signals. Pentoo Linux provides the most direct access to HackRF and toolboxes. Install Pentoo Linux on a separate drive, then you can use osmocom_siggen from a terminal to transmit signals such as near-field GSM bursts, which will only be detectable within a meter.









Installation of MGWin and cmake: Download and install the following packages:
- MinGW Setup (Go to the Installer directory and download setup file)
- CMake (I am using CMake 3.2.2 and I installed it in C:\CMake, this path is important in the commands we must send in the MinGW shell)
Download and extract the packages respectively in the path C:\MinGW\msys\…

Example: Beeswarm plot in R

library(foreign)

data <- read.dta("C:/Users/hellmund/Documents/MyStataDataFile.dta")

names(data)

install.packages('beeswarm')

library(beeswarm)

levels(data$group)

png(file="C:/Users/hellmund/Documents/il6.png", bg="transparent")

beeswarm(data$il6~data$group,data=data, method=c("swarm"),pch=16,pwcol=data$Gender,xlab='',ylab='il6',ylim=c(0,20))

legend('topright',legend=levels(data$Gender),title='Gender',pch=16,col=2:1)

boxplot(data$il6~data$group, data=data, add = T, names = c("","",""), col="#0000ff22")

dev.off()

Real world split-plot designs

Google Earth picture from a blog on statistics. A real world example near Christchurch (NZ) of a split-plot design. Today things have completely changed on location as the forest has grown considerably. Google Earth coordinate link.