我尝试使用函数glm和manyglm找到最佳的step模型：
glm.0 <- manyglm(rasy_spp ~+1, family="binominal")
glm.scope <- ~kytky+mechy+reka
step(glm.0,glm.scope)
结果如下：
> Start: AIC=745.03 rasy_spp ~ +1
>
> Df AIC
> + reka 26 730.39 <none> 745.03
> + mechy 156 891.33
> + kytky 689 1923.98
>
> Step: AIC=730.39 rasy_spp ~ reka
>
> Df AIC <none> 730.39
> - reka 26 745.03
> + mechy 156 918.22
> + kytky 689 1975.98
> **Error in data.frame(Step = I(change), Df = ddf, Deviance = dd, `Resid. Df` = rdf, : arguments imply differing number of rows: 2,
> 25, 13**
第一步结束后，计算结果如下：
glm.0 <- manyglm(rasy_spp ~+1, family="binominal")
glm.scope <- ~kytky+mechy
step(glm.0,glm.scope)
> Start: AIC=745.03 rasy_spp ~ +1
>
> Df AIC <none> 745.03
> + mechy 156 891.33
> + kytky 689 1923.98
>
> Call: manyglm(formula = rasy_spp ~ +1, family = "binominal") [1]
> "binomial(link=logit)"
>
> Degrees of Freedom: 138 Total (i.e. Null); 138 Residual
>
> (...) 2*log-likelihood: -20.94 Residual Deviance: 20.94 AIC: 22.94
> Warning message:
> **In data.frame(Step = I(change), Df = ddf, Deviance = dd, `Resid. Df` = rdf, : row names were found from a short variable and have been discarded**
我认为第一个示例中的错误消息与第二个示例中的警告消息有关...请如何解决此问题？ 谢谢！