Scribd
Upload
15 views12 pages

4.8 Slides - Example Melanoma Mortality (Count)

GLMM8

Uploaded by

Gabi Si Florin Jalencu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
15 views12 pages

4.8 Slides - Example Melanoma Mortality (Count)

GLMM8

Uploaded by

Gabi Si Florin Jalencu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

Example: Melanoma

Mortality (count)
Eventueel ondertitel
Cas Kruitwagen
Example case: Melanoma Mortality
• Data: Mmmec
library(mlmRev)
data(Mmmec)
?Mmmec

• Malignant Melanoma Mortality in the European Community


associated with the impact of UV radiation exposure.

2
Example case: Melanoma Mortality
• Data: Mmmec
data frame with 354 observations on the following 6 variables:

o nation - a factor with levels Belgium, W.Germany, Denmark,


France, UK, Italy, Ireland, Luxembourg, and Netherlands -> level 3
o region - Region ID - a factor. -> level 2
o county - County ID - a factor. -> level 1
o deaths - Number of male deaths due to MM during 1971–1980
-> outcome (number of deaths within county)
o Expected - Number of expected deaths due to MM. -> measure
for exposure (based on total number of deaths and person years
at risk, used as offset variable).
o uvb - Centered measure of the UVB dose reaching the earth's
surface in each county -> level 1 covariate

3
Example case: Melanoma Mortality
Examine the dataset
> Mmmec[1:4,]
nation region county deaths expected uvb
1 Belgium 1 1 79 51.2220 -2.9057
2 Belgium 2 2 80 79.9560 -3.2075
3 Belgium 2 3 51 46.5169 -2.8038
4 Belgium 2 4 43 55.0530 -3.0069

> Mmmec[301:304,]
nation region county deaths expected uvb
301 Italy 66 302 5 8.2140 6.0751
302 Italy 66 303 11 7.1600 6.6938
303 Italy 67 304 13 13.6230 1.2744
304 Italy 67 305 15 13.9220 1.6140

4
Example case: Melanoma Mortality
Some descriptives

> as.data.frame(table(Mmmec$nation)) #table in nice format


Var1 Freq
1 Belgium 11
2 W.Germany 30
3 Denmark 14
4 France 94
5 UK 70
6 Italy 95
7 Ireland 26
8 Luxembourg 3
9 Netherlands 11

> length(unique(Mmmec$region)) #number of regions


[1] 78

5
Example case: Melanoma Mortality
Some more descriptives

> summary(Mmmec$deaths)
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00 8.00 14.50 27.83 31.00 313.00

> summary(Mmmec$expected)
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.69 11.02 18.76 27.80 34.39 258.90

> summary(Mmmec$uvb)
Min. 1st Qu. Median Mean 3rd Qu. Max.
-8.900000 -4.158000 -0.886400 0.000204 3.276000 13.360000

6
Example case: Melanoma Mortality
• Let’s think about the analysis
• Deaths in county (count) -> Poisson regression
• Counties (=level 1) within regions (sample of regions in EU,=level 2)
• Predictor: UVB dose
• Random intercept per region?
• Random slope for UVB per region?

7
Example case: Melanoma Mortality
• Histogram of the outcome variable
> hist(Mmmec$deaths, xlim = c(0, 320), breaks = 320)

8
Example case: Melanoma Mortality
• Expected deaths -> Use offset in Poisson model

 E (deathsi ) 
ln  = β 0 + β1 X 1i + ... + β p X pi
 expectedi 

ln (E (deathsi ) ) − ln (expectedi ) = β 0 + β1 X 1i + ... + β p X pi

ln (E (deathsi ) ) = 1* ln (expectedi ) + β 0 + β1 X 1i + ... + β p X pi

9
Example case: Melanoma Mortality
• Some possible models
o Fixed effect only:
ln 𝐸𝐸 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖 = 𝑙𝑙𝑙𝑙 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑖𝑖 + 𝛽𝛽0 + 𝛽𝛽1 𝑢𝑢𝑢𝑢𝑢𝑢𝑖𝑖

o Random intercept per region:


ln 𝐸𝐸 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑗𝑗 = 𝑙𝑙𝑙𝑙 𝐸𝐸 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑖𝑖𝑖𝑖 + 𝛽𝛽0 + 𝑏𝑏0𝑖𝑖 + 𝛽𝛽1 𝑢𝑢𝑢𝑢𝑢𝑢𝑖𝑖𝑖𝑖

o Random intercept + random slope of UVB per region:


𝑙𝑙𝑙𝑙 𝐸𝐸 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑖𝑖 = ln 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑖𝑖𝑖𝑖 + 𝛽𝛽0 + 𝑏𝑏0𝑖𝑖 + 𝛽𝛽1 + 𝑏𝑏1𝑖𝑖 𝑢𝑢𝑢𝑢𝑢𝑢𝑖𝑖𝑖𝑖

10
Example case: Melanoma Mortality
Poisson regression model for deaths, regressed on a main effect of uvb,
and including a random intercept for region

> pmod1 <- glmer(deaths ~ uvb + (1|region), family = poisson, data


= Mmmec, offset = log(expected))

11
Example case: Melanoma Mortality
> pmod1
Generalized linear mixed model fit by the Laplace approximation
Formula: deaths ~ uvb + (1 | region)
Data: Mmmec
AIC BIC logLik deviance
661.4 673 -327.7 655.4
Random effects:
Groups Name Variance Std.Dev.
region (Intercept) 0.16968 0.41192
Number of obs: 354, groups: region, 78

Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.138601 0.049330 -2.810 0.004959 **
uvb -0.034434 0.009734 -3.538 0.000404 ***

12

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy