2

I tried a poisson regression model with the count of death as outcome and number of mutations in 4 different genes (G, O, S and V) as predictors and log (number of cases) as offset. The following is result of the model. Can someone kindly help me how to interpret this result? From the Estimate, can I say the mutation in Gene V is more harmful and mutation in S is least harmful. 

Call:
glm(formula = death ~ G + O + S + V + offset(log(cases)), family = poisson, 
    data = tt)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-64.907  -17.895  -10.481   -3.875   73.637  

Coefficients:
              Estimate Std. Error  z value Pr(>|z|)    
(Intercept) -2.807e+00  6.514e-03 -430.898   <2e-16 ***
G            6.913e-03  6.356e-04   10.876   <2e-16 ***
O           -9.373e-04  9.107e-05  -10.292   <2e-16 ***
S           -1.479e-03  1.688e-04   -8.762   <2e-16 ***
V            3.433e-02  1.947e-03   17.636   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

```
arshad
  • 863
  • 1
  • 7
  • 13

1 Answers1

3

For your poisson glm, you are modeling the death rate, that is the predicted rate is:

enter image description here

Starting with your intercept, the mean death rate is exp(-2.807e+00) = 0.06038588 and for example 1 unit of G (1 mutation guess) increases this rate by exp(6.913e-03) = 1.006937 times.

To answer your question, 1 mutation in Gene V for example increases the death rate 1.034926 fold.. I don't know if you can necessarily make the jump about it being harmful etc, that's your call.

Not sure what the average number of mutations in your data (which I suspect to be > 1), but will be useful to make sense of your coefficients in terms of that, for example, every 10 mutations increases the rate by .. fold

StupidWolf
  • 5,077