Why use the log_e function as link function for Poisson data in a GLM model
Clash Royale CLAN TAG#URR8PPP
up vote
1
down vote
favorite
This has been puzzling me for quite some time.
Suppose I have som data, that I have justefied should be described using a GLM model with a poisson as response.
As for choice of link function, You would normally go for the $log_e $function, and I understand that to be the case. Since we ofte observe, when working with count data. That the effekt of an observation is proportional to the side of size of the observation (ie. The predicter should be multiplicatice)
But would all this be true even if i chose log function such as the $ log_10$. Why do we always turn to the natural logarithm?
statistics
asked Aug 23 at 5:14
Viktor Jeppesen
15010
add a comment |Â
up vote
1
down vote
favorite
This has been puzzling me for quite some time.
Suppose I have som data, that I have justefied should be described using a GLM model with a poisson as response.
As for choice of link function, You would normally go for the $log_e $function, and I understand that to be the case. Since we ofte observe, when working with count data. That the effekt of an observation is proportional to the side of size of the observation (ie. The predicter should be multiplicatice)
But would all this be true even if i chose log function such as the $ log_10$. Why do we always turn to the natural logarithm?
statistics
asked Aug 23 at 5:14
Viktor Jeppesen
15010
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35
add a comment |Â
up vote
1
down vote
favorite
up vote
1
down vote
favorite
This has been puzzling me for quite some time.
Suppose I have som data, that I have justefied should be described using a GLM model with a poisson as response.
As for choice of link function, You would normally go for the $log_e $function, and I understand that to be the case. Since we ofte observe, when working with count data. That the effekt of an observation is proportional to the side of size of the observation (ie. The predicter should be multiplicatice)
But would all this be true even if i chose log function such as the $ log_10$. Why do we always turn to the natural logarithm?
statistics
asked Aug 23 at 5:14
Viktor Jeppesen
15010
This has been puzzling me for quite some time.
Suppose I have som data, that I have justefied should be described using a GLM model with a poisson as response.
As for choice of link function, You would normally go for the $log_e $function, and I understand that to be the case. Since we ofte observe, when working with count data. That the effekt of an observation is proportional to the side of size of the observation (ie. The predicter should be multiplicatice)
But would all this be true even if i chose log function such as the $ log_10$. Why do we always turn to the natural logarithm?
statistics
asked Aug 23 at 5:14
Viktor Jeppesen
15010
edited Aug 23 at 5:35
asked Aug 23 at 5:14
Viktor Jeppesen
15010
asked Aug 23 at 5:14
Viktor Jeppesen
15010
asked Aug 23 at 5:14
Viktor Jeppesen
15010
15010
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35
add a comment |Â
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35
add a comment |Â
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2891723%2fwhy-use-the-log-e-function-as-link-function-for-poisson-data-in-a-glm-model%23new-answer', 'question_page');
);
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
What difference would that make? The coefficients would change by a scaling factor of $ln(10)$ (and their standard errors would change by the same factor, so the p-values would be unaltered). Meanwhile the fitted values would all be exactly the same. You could take any present output and convert it to a log-10 link after the fact.
– Glen_b
Aug 25 at 3:35