p**5
发帖数: 2544 | 1 看了一篇文章，好像multiple imputation is very powerful | n****t
发帖数: 182 | 2 Yes, but MI has an assumption of MAR.
Because it is missing, you will need some assumptions one way or the other.
So a sensitivity analysis will be needed. | p**5
发帖数: 2544 | 3 I ran EM test (expectation maximization) and it rejected the H0, which means
my data is not MCAR.
But MAR cannot be tested, isn't it? So how people conclude that their data
is MNAR, but not MAR?
I heard although people have known about sensitivity analysis for quite some
time, no program can do it until the recent SAS 9.4 (proc MI) . Is this
true?
.
【在 n****t 的大作中提到】
: Yes, but MI has an assumption of MAR.
: Because it is missing, you will need some assumptions one way or the other.
: So a sensitivity analysis will be needed.
| n*******t
发帖数: 1296 | 4 这个是assumption吧，难道还有test来检验是MCAR还是MAR?
means
some
【在 p**5 的大作中提到】
: I ran EM test (expectation maximization) and it rejected the H0, which means
: my data is not MCAR.
: But MAR cannot be tested, isn't it? So how people conclude that their data
: is MNAR, but not MAR?
: I heard although people have known about sensitivity analysis for quite some
: time, no program can do it until the recent SAS 9.4 (proc MI) . Is this
: true?
:
: .
| n*******t
发帖数: 1296 | 5 你去找一本Frank Harrell的Regression Modeling Strategies看一下，里面讲的关于
怎么分析missing data很多东西很实用。用R。
【在 p**5 的大作中提到】
: 看了一篇文章，好像multiple imputation is very powerful
|
|
