<font face="trebuchet ms,sans-serif">Alexandre,<br><br>Deviance é uma "soma de quadrados" e não um quadrado médio. Então você não precisa multiplicar pelo grau de liberdade. Veja<br><br><span style="font-family: courier new,monospace;">y <- rnorm(20)</span><br style="font-family: courier new,monospace;">
<span style="font-family: courier new,monospace;">m0 <- lm(y~1)</span><br style="font-family: courier new,monospace;"><span style="font-family: courier new,monospace;">m1 <- glm(y~1, family=gaussian)</span><br style="font-family: courier new,monospace;">
<span style="font-family: courier new,monospace;">deviance(m0)</span><br style="font-family: courier new,monospace;"><span style="font-family: courier new,monospace;">deviance(m1)</span><br style="font-family: courier new,monospace;">
<span style="font-family: courier new,monospace;">summary(m0)$sigma^2*df.residual(m0)</span><br style="font-family: courier new,monospace;"><span style="font-family: courier new,monospace;">summary(m1)$deviance</span><br style="font-family: courier new,monospace;">
<br>Essas coisas de R² são complicadas. Pessoas da velha escola (ou não) sempre adorar uma medida em porcentagem. Deve trazer um certo conforto, sei lá. Penso que você deva argumentar outros aspectos da análise do que tentar justificar que tá tudo ok porque uma espécia de R² foi alto (ou não). De qualquer forma, uma media seria a porcentagem de deviance do modelo final contra o modelo nulo, algo como<br>
<br><span style="font-family: courier new,monospace;">x <- 1:30</span><br style="font-family: courier new,monospace;"><span style="font-family: courier new,monospace;">y <- rpois(x, x/4)</span><br style="font-family: courier new,monospace;">
<span style="font-family: courier new,monospace;">m0 <- glm(y~1, family=poisson)</span><br style="font-family: courier new,monospace;"><span style="font-family: courier new,monospace;">m1 <- glm(y~x, family=poisson)</span><br style="font-family: courier new,monospace;">
<span style="font-family: courier new,monospace;">1-deviance(m1)/deviance(m0)</span><br style="font-family: courier new,monospace;"><br>À disposição.<br>Walmes.<br><br clear="all"></font><span style="font-family: trebuchet ms,sans-serif;">==========================================================================</span><br style="font-family: trebuchet ms,sans-serif;">
<span style="font-family: trebuchet ms,sans-serif;">Walmes Marques Zeviani</span><br style="font-family: trebuchet ms,sans-serif;"><span style="font-family: trebuchet ms,sans-serif;">LEG (Laboratório de Estatística e Geoinformação, 25.450418 S, 49.231759 W)</span><br style="font-family: trebuchet ms,sans-serif;">
<span style="font-family: trebuchet ms,sans-serif;">Departamento de Estatística - Universidade Federal do Paraná</span><br style="font-family: trebuchet ms,sans-serif;"><span style="font-family: trebuchet ms,sans-serif;">fone: (+55) 41 3361 3573</span><br style="font-family: trebuchet ms,sans-serif;">
<span style="font-family: trebuchet ms,sans-serif;">VoIP: (3361 3600) 1053 1173</span><br style="font-family: trebuchet ms,sans-serif;"><span style="font-family: trebuchet ms,sans-serif;">e-mail: <a href="mailto:walmes@ufpr.br" target="_blank">walmes@ufpr.br</a></span><br style="font-family: trebuchet ms,sans-serif;">
<span style="font-family: trebuchet ms,sans-serif;">twitter: @walmeszeviani</span><br style="font-family: trebuchet ms,sans-serif;"><span style="font-family: trebuchet ms,sans-serif;">homepage: <a href="http://www.leg.ufpr.br/%7Ewalmes" target="_blank">http://www.leg.ufpr.br/~walmes</a></span><br style="font-family: trebuchet ms,sans-serif;">
<span style="font-family: trebuchet ms,sans-serif;">linux user number: 531218</span><br style="font-family: trebuchet ms,sans-serif;"><span style="font-family: trebuchet ms,sans-serif;">==========================================================================</span><br>