l*******f
发帖数: 243 | 1 手头上一个数据，y有很明显的seasonality pattern,但我有independent variables.
想法是先对y fit 一个time series 模型，然后对剩余的white noise fit linear
regression, 不知道有没有人这样做吗？可行吗？ 多谢 |
|
P*******9
发帖数: 9700 | 2 why not just adding a season dummy in the regression?
. |
|
d*****n
发帖数: 31 | 3 That's a good point,
But for the regression part we are assuming all Y's are independent, for the time series part, we are assuming all Y's are correlated, I do not know how to explain that |
|
m*e
发帖数: 146 | 4 什么情况下需要把regression的系数设成random的呢？和anova 中的random effect是
一回事吗？thanks.. |
|
c*********d
发帖数: 218 | 5 我以前以为要做negative binomial regression,independent variable必须有好多个 |
|
l***o
发帖数: 22 | 6 Thanks for replying.
I guess GLMSELECT is for "general linear model". Would you show me any
reference of how to use it for logistic regression? Sorry for stupid
question.... 3x!! |
|
P****d
发帖数: 113 | 7 I do not have any idea about this function or package.
But logistic regression is a special case of GLM.
So maybe you can specify "link function" to be "logit"
and "distribution" to be "binary", then the GLM is logistic. |
|
c*********d
发帖数: 218 | 8 怎么用R做 negative binomial regression啊？
急用 |
|
s***r
发帖数: 1121 | 9 dependent variable: b1, b2, p1, p2, m1, m2
independent variable: e1, e2, r1, r2, f1, f2
need to run 48 regressions, like
b1 = e1 r1 f1
b1 = e1 r1 f2
b1 = e1 r2 f1
b1 = e1 r2 f2
b1 = e2 r1 f1
b1 = e2 r1 f2
b1 = e2 r2 f1
b1 = e2 r2 f2
b2 = e1 r1 f1
b2 = e1 r1 f2
b2 = e1 r2 f1
b2 = e1 r2 f2
b2 = e2 r1 f1
b2 = e2 r1 f2
b2 = e2 r2 f1
b2 = e2 r2 f2
then do p1 as dependent variable
then do p2 as dependent variable
then do m1 as dependent variable
then do m2 as dependent variable
The SAS code is like:
p |
|
f*********8
发帖数: 165 | 10 请教logistic regression的 independent variable是categorical ariable 时，必须
是ordinal categorical 吗？
如果只是一般的categorical ariable，coefficent 的意义是什莫啊（如果有意义的话）
多谢。
在网上看到下面的这个例子。
http://nlp.stanford.edu/~manning/courses/ling289/logistic.pdf
里面的independent ariable 好像就是普通的categorical ariable，
比如说，ariable “cat” 的值是d,m,n,v; 另一个categorical ariable "follows"的值是P,V。 coefficent 的意义该怎末解释啊？
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.31827 0.12221 -10.787 < 2e-16
catd -0.16931 0.10032 -1.688 0.0 |
|
P****D
发帖数: 11146 | 11 看花眼了……我以为他每七年要做一个regression……1961-1967，1968-1975……
那就得用where statement了。 |
|
d*******o
发帖数: 493 | 12 1. Dimension reduction: run PCA/FA before the logistic regression.
2. Heuristic method: include all variables in the logistic model. Most
procedure has subset selection algorithm (backward/forward/stepwise),which
can decide the most significant variables. |
|
a*****3
发帖数: 601 | 13 请教有在database marketing行业做过的谈谈如何做customer segmentation/
profiling ? 我知道的方法有：
1.customer RFM
2.customer Life time value method
3.logistic regression
这三种方法有何优缺点？ 平时大家在工作中如何做segmentation的？
答案有一定信息的，有伪币答谢。 |
|
v*******o
发帖数: 10 | 14 Please give me some suggestions!!! Many Many Thanks!!!!
The problem is to investigate how the occurrence of an event (happened or
not happened) is relevant to some unknown factors.
Some of the unknown factors are numbers, some are characters (such as A, V,
PA,...).
I am thinking using logistic regression. For the occurrence of an event I
sent to 0 (not happened) and 1 (happened). For those non-numeric factors, I
will convert them to numbers (assign each one a number, for example: A->1, V
->2, PA... 阅读全帖 |
|
P****D
发帖数: 11146 | 15 What software do you use to fit your logistic regression model(s)? |
|
a*******k
发帖数: 25 | 16 各位高手，ordered probit regression, 请问如何写SAS程序得到 Pseudo R^2?
谢谢！！ |
|
d******o
发帖数: 59 | 17 see paper:
http://www.stat.lsu.edu/faculty/marx/CollinCSDA.pdf
Collinearity diagnostics for logistic regression can use condition index
with variance decomposition methods.
I have a paper discussing about VIF for GLMs It hasn't been published yet. |
|
o****o
发帖数: 8077 | 18 我觉得Ruppert等人那本semiparametric regression写得很浅显易懂,amazon上50多，
也不贵 |
|
d******3
发帖数: 93 | 19 看到几篇文献，做subic spline regression并且plot了predicted value和95%
confidence interval (CI)， 但是整条CI曲线上都没有referent
这是怎么回事？谁有好的讲义或tutorial么？谢了 |
|
|
k******w
发帖数: 269 | 21 I have four different regression equations with correlated errors.
how to estimate them in SAS/R/STATA? Thanks.
or some reference on internet will be appreciated. |
|
s*******r
发帖数: 769 | 22 In least squares regression, R2 alone cannot be used as a meaningful
comparison of models with different numbers of independent variables.
For a meaningful comparison between two models, an F-test can be performed
on the residual sum of squares, similar to the F-tests in Granger causality.
who told you "如果Rquare大于一个数，好像是0。7，那么这个model就fit data well
"?
如果Rquare大于一个数，好像是0。7，那么这个model就fit data well。 |
|
j******1
发帖数: 62 | 23 Is there a way to do crossvalidation for logistic regression in SAS?
Thanks a lot! |
|
d*******1
发帖数: 854 | 24 大家好， 有一个简单的regression model:
Y=b1X1+b2X2+b12X1*X2+e. (Y是心率， X1血压， X2体温）
这里有个interaction term X1*X2. 如果X1X2一个categorical一个continuous的话，
这个interaction coefficent 很好解释， 比如说血压和心率的关系和性别有关。
但是如果X1X2都是continuous, 比如是血压和体温。 那么如何解释血液和体温的
interaction呢？ |
|
a*******k
发帖数: 25 | 25 Under ordered probit regression, when the DV is defined into 0 to 5, how to
explain the results of marginal effects?
Thanks a lot! |
|
w*********8
发帖数: 70 | 26 假设有组binary data，比如说不同年龄人群里面的男女分布。现在我们把性别0，1分
布转化：在25岁的人群里面，n=100, 75female, 那75/100=0.75是我们的第一个值；在
26岁人群里，n=200, 100 female,那100/200是第二个值 and so on....这样的话我们
得到的这组数据就是continuous了。问：这时候可以用linear regression来处理这组
数据吗？首先，做过normality test了，satisfied. 其他几个assumption也满足。
多谢！ |
|
k*******a
发帖数: 772 | 27 不过你怎么解释你的model？
linear regression的y取值范围不可能限制与(0,1) |
|
a****r
发帖数: 1486 | 28 这跟logistics regression有啥太大区别吗？ |
|
P****D
发帖数: 11146 | 29 Applied Regression Analysis and Multivariable Methods (Duxbury Applied) by
David G. Kleinbaum, Lawrence L. Kupper, Azhar Nizam, Keith E. Muller
这书好像有人特喜欢有人特讨厌。我觉得还行。 |
|
c*******n
发帖数: 300 | 30 y|X是normal. linear regression并不要求error follow normal distribution.但是
，常见的估计方法却需要这个假设。 |
|
D*********2
发帖数: 535 | 31
nope，是Y normal。
如果X是fixed variable, Y|X没意义。
如果X是random variable，其实在simple linear regression的数理推理中需要的是（
Y,X） bivariate normal. |
|
d******e
发帖数: 7844 | 32 应该把你打回去重学。
如果是random design, linear regression的要求是，E(Y|X)是linear function of X
和Var(Y|X)是normal。更一般的要求是monoment condition, Var(Y|X) = 0和Var(Y|X)
joint normal只是一个特例罢了 |
|
D*********2
发帖数: 535 | 33 重学回来了，那个jt normal确实是俺的错～ 反正人也丢了，干脆继续，再请教一下啊～
Fixed Effect应该没错吧，error~ Normal, Y ~ Normal, 所以in practice, 就直接测
Y。
Random Effect
1. in practice 这么验证 normality? Y|X 具体咋implement?
2. 我看wiki上其实写的是 error|X 的一系列 assumption
楼主，great question! 但您问的是simple linear regression呢，还是ordinary
least
square?
X
X) |
|
f**********t
发帖数: 1001 | 34 我问的是ordinary least square，不是simple linear regression
顶一下你问的random effect问题。呵呵
一下啊～
practice, 就直接测
ordinary |
|
D*********2
发帖数: 535 | 35 ...那您让兄弟姐妹们回答了一圈simple linear regression的问题。
OLS的假设应该就只是关于error term的，or, to be specific, E[\epsilon|X]=0,
Var[\epsilon|X]=\sigma^2 I_n。normality 都不用。 |
|
c****n
发帖数: 271 | 36 Applied Logistic Regression, 2nd Edition
By: David W. Hosmer
By: Stanley Lemeshow
Should anybody share this ebook with me?
Thanks a lot |
|
|
F****n
发帖数: 3271 | 38 I hope I could, that would definitely make things much easier. But I don't
think I can do that in this case, as it would be completely meaningless.
Moreover, even you can do so, you get a proportional odds model, which
assume the increase of cumulative probability are uniform across different
ranks.
In multinomial regression, on the contrary, you have one set of
coefficients for each category compared to the reference. And that's what I
want.
? |
|
F****n
发帖数: 3271 | 39 Actually the results are not bad. Technically it is not difficult. I just
feel a little nervous about the potential flaw in this approach.
Basically my dependent has 4 categories, and I have run the regression 3
times, each with a different reference category. |
|
p***l
发帖数: 1775 | 40 多谢楼上回帖，很多书上涉及ridge regression的时候，讲的很粗略。
还有我第一个问题，用stepwise或者其他手段做变量筛选后，检测剩余变量的vif，发
现依然大于10。（center他们也没有用）这时候转向ridge，make sense吗？还有其他
补救措施吗？ |
|
n******e
发帖数: 476 | 41 唔，是 logistic Joinpoint Regression。这个 R 里面有，可是还是不会用啊。比较
几个年份的 weighted frequency，看有没有 increase or decrease trend，不同的
treatment 在这几年的 trends 有没有 sig. difference。
不知道怎么弄，怎么知道有多少个 knots 呢。有用过的么，给点入门介绍吧。 |
|
n******e
发帖数: 476 | 42 看了那个 Joinpoint free software，明白多了。R 里那个 ljr 就是把人家那个
software 给简单做了一下，要用的话要去看人家的 manual 比较好。
比较 trends,老板说其实只要比较一段就可以。那就是比较两个 linear regression
了，用个 interaction 看 p-value 吧，我想。 //擦汗擦汗 |
|
e*******e
发帖数: 1144 | 43 X is an n*p matrix with n observations and p variables, \beta is the p*1 vector for a regression model, and then Y = X \beta is the n*1 prediction vector.
My question:
Define diag(\beta) as the p*p diagonal matrix whose diagonal elements are \beta, then is there a name for the quantity: X diag(\beta)? Note that this quantity is a n*p matrix, where each column is basically "the prediction values from one dimension of \beta".
Thanks! |
|
c*t
发帖数: 1063 | 44 【 以下文字转载自 Biology 讨论区 】
发信人: cxt (cxt), 信区: Biology
标 题: 请教：multilevel logistic regression?
发信站: BBS 未名空间站 (Tue Feb 8 23:38:16 2011, 美东)
什么样的情况用这个分析数据？用这个有什么好处？有什么替代的分析方法么？
万分感谢！！ |
|
y********9
发帖数: 13 | 45 急问高手,怎样在SAS实现logistic regression里independent variable重要性排序？
（所有的independent variable都significant,P-value all <0.0001.但想知道哪个
variable最重要，哪个最不重要，需要按importance ranking score排序）。多谢！！
！ |
|
p******r
发帖数: 1279 | 46 Is zero-inflated model only valid for poisson family?
what i'm doing is Ordinal regression. it the zero-inflated model still ok to
do this? thanks! |
|
A*******s
发帖数: 3942 | 47 like what i said b4,
"i think poisson's nature is about counts or time length. not sure if other
types of ordinal data could fit in this approach. "
would like to know Niu Ren's opinion. show what kind of ass u r.
btw, google "poisson regression depression score", some results come up. |
|
G*****m
发帖数: 222 | 48 1。ols=mle?
Normality (?). It is sometimes additionally assumed that the errors have
normal distribution conditional on the regressors:[4]
see：
http://en.wikipedia.org/wiki/Ordinary_least_squares
2。如果ppl不愿意接收他们 highly depressed的solution:
OLS, 得到error term。plot error against score.regress error on score, 检测是
否相关。解决法子，我也不清楚...要看你的X， literature。不过好像bootstrapping
通吃？
啊！ |
|
d*******o
发帖数: 493 | 49 if y is continuous, x is discrete, then one-way anova;
if y and x are both continuous, then linear regression. |
|
T*******I
发帖数: 5138 | 50 从当前的方法论系统来看，Da Shagen, Dapangmao, juniorstat和prior都给出了最恰
当的建议。LZ的问题应该不是经典的linear regression, 而是基于One-way方差分析上
的广义线性模型。但powerpower可能有自己的理解。建议powerpower多费点笔墨谈谈那
个indicator variable是怎么构造的。谢了。 |
|
