j*******g
发帖数: 33 | 1 正在 fitting a model, right now tried GLM with Poisson and Negative binomial
distribution, the residual plots still sucks. I am thinking zero inflated
NB with count data
Hope I can get some suggestion to deal with scew, 0, and outliers.
Thanks |
l***a
发帖数: 12410 | 2 outliers难道不先直接去掉？
binomial
【在 j*******g 的大作中提到】
: 正在 fitting a model, right now tried GLM with Poisson and Negative binomial
: distribution, the residual plots still sucks. I am thinking zero inflated
: NB with count data
: Hope I can get some suggestion to deal with scew, 0, and outliers.
: Thanks
|
j*******g
发帖数: 33 | 3 thx
binomial
【在 j*******g 的大作中提到】
: 正在 fitting a model, right now tried GLM with Poisson and Negative binomial
: distribution, the residual plots still sucks. I am thinking zero inflated
: NB with count data
: Hope I can get some suggestion to deal with scew, 0, and outliers.
: Thanks
|
j*******g
发帖数: 33 | 4 现在还没有对数据作任何处理，
想试ZINB，ZIP二选一，outlier的话是把不fit的点头和尾都砍掉？想敲定一个model fit差不多的时候再考虑kick out outliers.
涉不涉及influencial points? 总觉得最后那个leverage~st.Pear resi看不懂
刚学着fitting实际问题，帮建议一下，谢谢! |
s*********e
发帖数: 1051 | 5 you are doing the wrong way for model diagnosis.
for count outcome model, you should compare the observed probability with
the predicted probability.
check my paper "Count Data Models in SAS" for details. |
j*******g
发帖数: 33 | 6 Thanks a lot, im reading your paper now.
【在 s*********e 的大作中提到】
: you are doing the wrong way for model diagnosis.
: for count outcome model, you should compare the observed probability with
: the predicted probability.
: check my paper "Count Data Models in SAS" for details.
|
l*********s
发帖数: 5409 | 7 cool!
【在 s*********e 的大作中提到】
: you are doing the wrong way for model diagnosis.
: for count outcome model, you should compare the observed probability with
: the predicted probability.
: check my paper "Count Data Models in SAS" for details.
|
