CODEWITHSUNDEEP
rintegerpoisson
S G
S G

Reputation: 21

Trouble with the poisson model-integer number

I'm having trouble with the running of a log-log regression model with Poisson. How should I prevent this warning message? It's also the first time that I use Poisson so I really don't know how to do. Thanks a lot

sardegnalog.lm <-glm(log1p(fulldata[381:400,1])~log1p(fulldata[381:400,2])+log1p(fulldata[381:400,3])+log1p(fulldata[381:400,4])+log1p(fulldata[381:400,8]), family="poisson")
Warning messages:
1: In dpois(y, mu, log = TRUE) : non-integer x = 8.868132
2: In dpois(y, mu, log = TRUE) : non-integer x = 9.885069
3: In dpois(y, mu, log = TRUE) : non-integer x = 9.410911 
4: In dpois(y, mu, log = TRUE) : non-integer x = 7.876259
5: In dpois(y, mu, log = TRUE) : non-integer x = 11.826326
6: In dpois(y, mu, log = TRUE) : non-integer x = 9.632728
7: In dpois(y, mu, log = TRUE) : non-integer x = 9.872616
8: In dpois(y, mu, log = TRUE) : non-integer x = 6.899723
9: In dpois(y, mu, log = TRUE) : non-integer x = 9.027379
10: In dpois(y, mu, log = TRUE) : non-integer x = 16.733528

summary(sardegnalog.lm)

Call:
glm(formula = log1p(fulldata[381:400, 1]) ~ log1p(fulldata[381:400, 
2]) + log1p(fulldata[381:400, 3]) + log1p(fulldata[381:400, 
4]) + log1p(fulldata[381:400, 8]), family = "poisson")

Deviance Residuals: 
Min      1Q  Median      3Q     Max  
-3.267  -2.082  -1.093   1.085   3.123  

Coefficients:
                        Estimate Std. Error z value Pr(>|z|)    
(Intercept)                 -17.5129     5.2594  -3.330 0.000869 ***
log1p(fulldata[381:400, 2])   1.3144     0.4632   2.838 0.004544 ** 
log1p(fulldata[381:400, 3])   0.7884     0.2384   3.307 0.000944 ***
log1p(fulldata[381:400, 4])  -0.1477     0.2613  -0.565 0.571836    
log1p(fulldata[381:400, 8])  -0.7765     0.2960  -2.623 0.008715 ** 
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for poisson family taken to be 1)

Null deviance: 144.602  on 19  degrees of freedom
Residual deviance:  80.231  on 15  degrees of freedom
AIC: Inf

Number of Fisher Scoring iterations: 6

Upvotes: 1

Views: 1535

Answers (1)

Glen_b
Glen_b

Reputation: 8252

It's difficult to be certain because you haven't been sufficiently explicit about the model, but it looks like you need to simply drop the log1p on the left hand side of the formula; the Poisson glm already has a log link function by default (but you don't need to add 1, because it's the mean, not the data that is transformed to the scale of the linear predictor).

You would still need it on the right though.

Upvotes: 0

Related Questions