Senhores,<br>estou ajustando uma regressão logistica via glm<br><br>na hora de validar meu modelo com amostra teste,no comando predict, de acordo com o Help o default para a previsão é o log do odd ratio e o type="response" retornaria as probabilidades,<br>
<br>porem me parece que o default está retornando as probabilidades e o response o ajuste linear.<br><br><br>segue o trecho do help<br><br><table summary="R argblock"><tbody><tr valign="top"><td><code></code><br></td>
<td>
the type of prediction required.  The default is on the
scale of the linear predictors; the alternative <code>"response"</code>
is on the scale of the response variable.  Thus for a default
binomial model the default predictions are of log-odds (probabilities
on logit scale) and <code>type = "response"</code> gives the predicted
probabilities.  The <code>"terms"</code> option returns a matrix giving the
fitted values of each term in the model formula on the linear predictor
scale.
</td></tr></tbody></table><br>Alguem sabe alguma coisa sobre isso?<br><br clear="all">Att.<br>-- <br>Vinicius Brito Rocha.<br>Estatístico e Atuário,<br>Mestre em Pesquisa Operacional<br><br><a href="http://www.aplicademic.blogspot.com" target="_blank">www.aplicademic.blogspot.com</a><br>
<a href="http://twitter.com/viniciusbritor" target="_blank">http://twitter.com/viniciusbritor</a><br><br>"Não se preocupe muito com as suas dificuldades em Matemática, posso assegurar-lhe que as minhas são ainda maiores." - Albert Einstein.<br>
<br><br>