由买买提看人间百态

topics

全部话题 - 话题: glm
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)
W****W
发帖数: 508
i***g
发帖数: 11
f*******n
发帖数: 70
3
来自主题: Statistics版 - SAS下载安装完全教程
就是我给的那个连接,不少人说装不了,俺负责到底,刚刚自己去下载完了
那个SAS文件,已经在roommate机器上装上了,Win XP home,没任何问题,run
了几个proc, GLM, ARIMA, 都OK. 下面就安装扫一下盲.
连接:
http://stat.ustc.edu.cn/data/software/sas8/SAS8.BIN
科大的ftp,还能放多久不知道,快点下吧.
btw,下面这个网站还有别的东东,能不能下不知道,反正这个SAS就是
在这儿找到的.
http://text88.myrice.com/downloads/stata/
安装步骤:
下完了是个.bin文件,可以刻盘或者用winiso解压后直接从硬盘上安装.
安装中途我碰到一个error,israel什么文件拷不了,skip掉.会重启一次
安装最后要initiate,找那个SETINIT文件,应该在V8/core/sasinst下面.
装完了把源盘(.bin文件)里的那个crack目录下的东西拷到V8/目录下覆盖掉.
然后我运行了一下,run SAS/BASE 没问题, 但别的package象SAS/
g**r
发帖数: 425
4
来自主题: Statistics版 - Re: 请问SAS如何学起?
系统全面的学SAS我觉得是浪费时间。SAS那么多的内容,光SAS GRAPH
就够“系统全面”的学上两个月的,投入产出比不大吧。其他的象SAS AF,
SCL, SAS QC 等等多是一辈子一般人也用不着的东西。
DATA STEP、MACRO、PROC IML、 PROC SQL除外,这些是工具性质,以及锦上添花的ODS,
如果经常用SAS,这些是要弄弄熟的。其他的,大概翻翻手册知道有这个东西就行了。比
如PROC REG,你知道有个PROC REG就行了,到做PROJECT的时候要用了,再去翻手册都来
的及。
我觉得学SAS最好的方法就是做PROJECT,那个是逼着你用,用多了就“系统”了。
我认为最好的参考书还是SAS MANUAL。那些个文字99%是废话,最好是说例子多。
好到平时SAS编程多数可以依葫芦画瓢。
当然了,如果是比较FANCY的统计模型,是自己一本书的。比如SURVIVAL ANALYSIS啦,
CATAGORICAL 啦, GLM 啦,PROC MIX啦,ETS啦,要系统得一个一个的来。
忽然有个想法,建议本版的大侠们有时间不妨总结一些小册子出来。
到时候出“统
z******j
发帖数: 1265
5
来自主题: Statistics版 - Re: About GEE models?
the major disadvantage of GEE is that it only takes care of 1 hierarchical
structure and makes inference on the mean level, whereas HGLM can handle
several-layer structure and makes inference on the independent cluster level.
for example, there are 2 treatments for two groups of ppl. each person within
the group are observed several time points. In this case, the GEE model makes
inference on the effect of treatments. the GLM makes inferene on the person level.
g**r
发帖数: 425
6
DATA STEP是必须要搞熟的,除了读入文件DO LOOP什么的,
retain, first. last.; date format,这些都不在话下
MACRO很重要,很重要很重要;但MACRO学起来是很容易的,比DATA STEP
容易好多
PROC TABULATE (or PROC REPORT);要搞搞熟的,这个很重要
Haworth的 PROC TEBULATE BY Example是值得花时间读一读的
我痛恨胡乱用PROC TABULATE的PROGRAMMER,那么强的工具,经常被人
糟塌--明明三五行就解决的事情,楞写出百十行来,让人看的那个难受。
直观易懂的PROC MEANS, PROC FREQ当然不用说了
PROC GPLOT和PROC GCHART要翻一翻的,AXIS,PATTERN, LENGEND,SYMBLO什么的
临时学倒是来的及。但人家问你的时候你要知道SAS GRAPH这回事情。
如果是搞TRIAL的,PROC LIFETEST, PROC PHREG要看一看的;
否则的话,PROC REG, PROG LOGISTIC, PROC GLM翻翻
p********a
发帖数: 5352
7
☆─────────────────────────────────────☆
kouzhong (jinzhongyue) 于 (Mon Jan 1 17:38:51 2007) 提到:
光SAS/STAT 手册就有7000多页, 一个GLM PROCEDURE 就有好几百页。东西是好,但太
多了,大家是怎么学习的?欢迎探讨!
祝大家新年快乐!
☆─────────────────────────────────────☆
papertigra (长工胖头猪) 于 (Mon Jan 1 17:41:50 2007) 提到:
那是大词典,不是用来学习的
☆─────────────────────────────────────☆
kouzhong (jinzhongyue) 于 (Mon Jan 1 17:43:21 2007) 提到:
多谢版主,这两天我没事在狂啃呢!
☆─────────────────────────────────────☆
comeandgo (春困秋乏夏打盹) 于 (Mon Jan 1 19:34:36 200
p********a
发帖数: 5352
8
来自主题: Statistics版 - [合集] excel + vba 做统计能行么?
☆─────────────────────────────────────☆
leftHand (Hold on...) 于 (Fri Jan 27 16:50:44 2006) 提到:
一个小公司不愿意掏钱买大软件,R 他们不懂,另外画图
他们也希望在excel里边画。所以想看看vba能不能应付的了
☆─────────────────────────────────────☆
sir ( 郎 ) 于 (Fri Jan 27 17:02:55 2006) 提到:
要看做什么了
excel能载入的数据量通常是不够的

☆─────────────────────────────────────☆
leftHand (Hold on...) 于 (Fri Jan 27 17:40:23 2006) 提到:
就是一般常干的那点东西了
相当于sas里base, graph的大部分功能, glm,reg,anova一类的就够了
☆─────────────────────────────────────☆
sir ( 郎 ) 于 (Fri Jan 27
c**********e
发帖数: 2007
9
来自主题: Statistics版 - How to make the SS3 a dataset?
In PROC GLM procedure, model y= x1 x2/ss3; will output the
SS3 table. But how to make the SS3 a SAS dataset? Thanks.
l*******f
发帖数: 243
10
来自主题: Statistics版 - 求教:几个模型的区别
大家好, 这学期刚上了一门博士的课程,关于统计计算的。觉得有些难。
求教大家:
LM (linear model), LMM(linear mixed model), GLM(generalized linear model)
, GLMM(generalized linear mixed model)之间的区别和联系, 谢谢大家的帮忙。
大家有好的这方面的书可以推荐的吗?谢谢啦。
j*****g
发帖数: 222
11
phase I dose range define
怎么处理missing value and outlier?
如果我用glm model
难道missing value直接不管,outlier直接删除么?
谢谢啦!
a********a
发帖数: 346
12
来自主题: Statistics版 - help sas code for choosing control group
data one;
input lake $ bacteria;
datalines;
Pristine 3933
Pristine 458
Pristine 94
Vacation 862
Vacation 437
Vacation 538
Pleasant 4481
Pleasant 276
Pleasant 4208
Sardine 7456
Sardine 4394
;
I want to test the mean of different lakes. So I used the following sas code,
proc glm data=one;
class lake;
model bacteria=lake;
run;
I know sas will treat 'Vacation' lake as the control group. If I want to use
'Pristine' as control group, how do I write in sas code
S***e
发帖数: 108
13
来自主题: Statistics版 - help sas code for choosing control group
data two;
set one;
if lake='Pristine' then group=4;
else if lake='Vacation' then group=1;
else if lake='Pleasant' then group=2;
else if lake='Sardine' then group=3;
else group=.;
run;
proc glm data=two;
class group;
model bacteria=group/solution;
run;
a********a
发帖数: 346
14
来自主题: Statistics版 - help sas code for choosing control group
Yes, creating dummy variables is a way to it. Is there any way to write such
like reference='Pristine' in proc glm? I know we can write ref='' in proc
logistic.
Thanks
m********1
发帖数: 368
15
I am wondering if there are any statements in PROC GENMOD (or any PROC that
can deal with generalized linear models) that can fit the model
g(\pi)=beta0+beta1*x
with the restriction
g(\pi_0)=beta0+beta1*x0
Here g() is the link function, x0 and \pi0 are already known. We want to
estimate beta0 and beta1.
I don't know whether some SAS statement can specify this kind of restriction
in fitting the generalized linear model. Would you give me some advice?
Thank you very much for your help!
m********1
发帖数: 368
16
Exactly! However, I didn't find any software that can specify this
restriction...

that
then
genmod to
specific model, you may be able to do it with other procedures. For example
, you can do proc logistic, if the link is a logit...
j*****e
发帖数: 182
17
Use Proc Nlmixed to write your own likelihood function.
m********1
发帖数: 368
18
Thank you very much. Your hint is very useful.
I have just noticed that, using PROC NLMIXED, in the "model" statement we
can use general(ll) to specify a general log likelihood function that we
construct using SAS programming statements.
Are there any other solutions for this problem?
m********1
发帖数: 368
19
I think the suggestion of jsdagre on the #4 floor above is better.
My original idea is exactly same as yours, however, it has a problem: \alpha
_0=g(\pi_0) is already given. So the only parameter to estimate is \alpha_1=
\beta_1.
To use fisher-scoring and find a solution to maximize the defined log likelihood is
more straight forward, I think. Only need to define the log likelihood function and PROC NLMIXED will solve it for you.
o****o
发帖数: 8077
20
不过这种方法你就可以试试offset的技巧with no intercept, 这样就不用自己写似然
方程了

alpha
1=
likelihood is
function and PROC NLMIXED will solve it for you.
m********1
发帖数: 368
21
有道理,多谢oloolo大人!我终于明白怎么设置offset了!
那么至少有两种办法可以做这个问题了。
g*********n
发帖数: 119
22
you can use R with GLM to achieve respective deviance for two models, then
use the chi square test with df=2 to test the hypothesis.
s***i
发帖数: 49
23
Can I use the "null deviance" in R as the deviance with 0 explanatory
variable? But how do I model a logit regression with 0 explanatory variable
in the first place, using GLM?
f***a
发帖数: 329
24
来自主题: Statistics版 - which nonparametric test shoud I do
Or just fit a LM or GLM to see what the coefficient of condition looks like?
s*****n
发帖数: 2174
25
来自主题: Statistics版 - 求Travelers Insurance的面试经验
曾经拿到过他家的三个offer, 不过都是full time.
一般会问一些基本的统计概念, 比如glm啊什么的.
还有就是一些实例估算问题, 比如让你估计一下美国
境内有多少辆车, NYC的居民一年要吃掉多少冰激凌.
如果是特别面试 Personal Lines 的话, 了解一下
汽车保险, 房屋保险什么的常识.
r***a
发帖数: 20
A*******r
发帖数: 194
27
来自主题: Statistics版 - on-site test都有什么问题?
我不在这个area, 不过这种工作 logistic regression 一定要熟,Interpretation 很
重要。 再扩展一点GLM。
t******g
发帖数: 1815
28
来自主题: Statistics版 - 对应STATA的SAS CODE
running this in STATA
glm Y X, link(log) family (nbinomial)
有没有相对应的SAS CODE?
谢谢
d*******1
发帖数: 293
29
来自主题: Statistics版 - 笨人求问sas function
第二个好像是GLM model 的题
b*******n
发帖数: 3
30
Hi,
I need to combine both group mean, std and withingroup p-value, ANOVA p-
value for a phase II efficacy summary table. I plan to use proc report, but
I am not sure how to pull p-value from proc GLM/MIXED. Any reference SAS
code or books?
Thanks a lot!
c*****m
发帖数: 4817
31
summary里有p value
c*********t
发帖数: 340
32
谢谢ls帮忙:)
我知道summary里有
可是我是想把每次的P值都输入到一个文件里
我要做六千多个logistic regression
不想一个一个地用SUMMARY看。。。。
c********w
发帖数: 4
33
来自主题: Statistics版 - Biostatistics Wanted-Job Openning
Great Job opportunities for Biostatistics wtihin Gloable leading CRO, in
Mainland China.
Job Description
• Ideally at least 3 years of experience with SAS programming (inc.
SAS/BASE, SAS/STAT, SQL, MACRO) and experience with large size databases
• Experience with statistical analysis and modeling techniques (eg,
ANOVA, GLM, regression, Bayesian models, cluster analysis)
• Experience with medical coding (eg, ICD-9, CPT-4)
• Experience in report creation using
s*********e
发帖数: 1051
34
you didn't miss anything. the answer given was wrong. x * x in model
statement won't work in proc reg.
however, try it in proc glm/genmod. i would be surprised if it won't work.
please post results after you try such that others could learn as well.
thanks.
z****k
发帖数: 1057
35
glm的话相应的语句应该怎么写?
我直接用model y = x1 + x2 + x1*x2
结果好像只给出ANONVA,没有coefficient
c*****m
发帖数: 4817
36
来自主题: Statistics版 - R question
when use glm, link=log
how to get the estimate and confidence interval for the "effect" of variable
,
i.e. the original effect or exp(coefficient) so that I can better interprate
the results
I know in SAS, I can use statement like
estimate "effect" income /exp;
Thanks
v*******g
发帖数: 334
37
proc glm ;
contrast ;
does it work?
L**********A
发帖数: 7115
38
来自主题: Statistics版 - proc GLM ; contrast : why this is QUADRATIC ?
google ORTHOGONAL POLYNOMIAL COEFFICIENTS
v*******g
发帖数: 334
39
来自主题: Statistics版 - proc GLM ; contrast : why this is QUADRATIC ?
Thanks, i did .
Ok they're COEFFICIENTs of ( A+B)^2 or (A+B+C) ^2
but why the it has QUADRATIC effect?
s*********o
发帖数: 42
40
如果要比较varialbe prog 的level 2 vs. level 3, 以及group 1 和 group 2 之间的
区别。prog 一共有3个level.
sas code如下:
proc logistic data = hsb2;
class prog/param=glm;
model hiwrite(event="1")=female read math prog;
contrast '1 vs 2 of prog' prog 1 -1 0/estimate;
run;
请问这里是怎样给prog做coding 的,为什么prog是1 -1 0, 为什么 contrast '1 vs 2
of prog' ,给出来的是1 vs 3 and 2 vs 3的odds ratio estimate? 谢谢。
s*****n
发帖数: 2174
41
来自主题: Statistics版 - 线性回归中log tranformation
所谓工业界里面希望求E(y), 无非就是人们已经习惯了"average"这个概念,
从而把这个概念应用到各个细节, 不管所谓的"average"到底合理不合理.
比如你说的例子, 想预测revenue, 但是只有log(revenue)才好用.
这种情况从纯统计来说, 根本就不应该来预测y, 只应该预测log(y), 或者
说人们看到revenue, 脑子里面就应该想成是log(reveune), 遇到reveune
就自动取log, 根本没有原始revenue的概念. 你的project中, 讨论的永远
是log(revenue), 根本不存在back-trans的问题. 当然这是理想情况, 现实中, 你未必
能说服你的manager把脑筋转过来. 这种情况下, 如果实在无法说服别人,
那就凑合transform回去好了, 不过麻烦的问题是, 你manager有可能问你,
"我自己把原数据取了个平均, 为什么和你给我fit的差这么远呢?" 这种情况
你还得费很大力气给他解释为什么 E(Y) =/= exp(E(log(Y))).
当然也有其他的方法, 比如用glm什么的. 只要结果能
q********i
发帖数: 795
42
来自主题: Statistics版 - 求助:hierarchical linear model 的问题
同意你后半部分的comment, 但是hierarchical linear model跟baysian没有必然联系
。很多multi-level random effects model都可以看成是hierarchical model,根本用
不着prior.一般proc mixed,甚至proc glm都能做。

data
o******6
发帖数: 538
43
☆─────────────────────────────────────☆
Bighappy (快乐大大大) 于 (Tue Feb 5 23:34:09 2008) 提到:
问题是这样的:Response是一些Score, Treatment Group就2个Level:Drug和Placebo
,性别为Covariate。我开始用的Two-way ANOVA,已经发现Treatment-by-gender是
Insignificant的了,本来是只想用Main Effects Model来说明Drug作用的,但客户的
要求将Treatment的作用分男女表述出来。现在有2个方案:
(1)将Insignificant Interaction还放在Model里:
proc glm data=scores;
class treat gender;
model scores = treat gender treat*gender;
estimate 'drug effect for males' treat 1 -1;
estimate 'drug effe
w*******a
发帖数: 2409
44
来自主题: Statistics版 - 问一个 classification 的问题
GLM
o******6
发帖数: 538
45
☆─────────────────────────────────────☆
sturtle (huhu) 于 (Sat Mar 1 12:16:54 2008) 提到:
我现在有个数据,形如:
a b c d
...data...
a,b,c,d是数据的column name, 我现在想做a vs d, b vs d, c vs d的proc mixed

但是非常奇怪的是proc mixed 不能像 proc glm那样用 model a b c = d 一次就得到
所有model,必须 model a=d, model b=d这样一个个的来, 所以我打算做个循环来自
动做这件事,但是我不知道怎么自动的读出column name和生成 model X=d这样的语句
,请大侠们帮忙。
☆─────────────────────────────────────☆
chrd (对酒当歌,杜绝灌水) 于 (Sat Mar 1 12:29:27 2008) 提到:
1 proc contents
w***y
发帖数: 114
46
for categorical
%macro cate(var)
proc freq data=
tables &var.*treatment/exact;
output out=pvalue(keep=XP2_fish) exact;
run;
%mend;
for continuous
%cont(var)
proc glm data= outstat=pvalue_con(keep=prob _type_ _name_ where=(type="
ss3"));
class treatment;
model &var.=treatment;
run;
%mend;
you need change it to macro sas code
m*******s
发帖数: 469
47
proc univariate
proc freq
proc glm
proc lifetest
proc phreg
proc mixed
真正弄懂玩透这几个SAS命令,药厂95%的project可以应付了。
s****y
发帖数: 21
48
来自主题: Statistics版 - 问个SAS output的问题
使用了proc mixed, 生成predicted value之后怎样把结果output成一个sas data set
,以方便以后使用predicted这个variable。
在其他proc,比如proc reg,proc glm中是不是做法也该差不多?
g******7
发帖数: 19
49
来自主题: Statistics版 - How to compare two parameters.
one possible solution is using the simutaneous equation modeling, or
structural equation modeling (the so-called SEM)
but first you will need to establish a theoretical frame/model. if you can,
it is very straitforward to test: coefficient of Z = coefficient of W by
model comparison (something similar to GLH test in GLM models).
In SAS, you use proc Calis
the other software to do it include: LISREL, AMOS etc.
w**********y
发帖数: 1691
50
来自主题: Statistics版 - 统计分析问题请教, Baozi question!
来抢包子...
" analyze those 5 variables all together"
请google: multivariate ANOVA..
简单的说,你这5个response variale如果是independent的,那很好办啦,就一个一个的
做ANOVA就行了(不过你要adjust significant level)..
而MANOVA就是把这5个response当作一个vector,然后去比较你这A,B,C,D的response
vector有没有significant different..而test用的方法基本都是非参的方法..
SAS里面实现MANOVA的例子:
PROC GLM DATA=RAWDATA;
CLASS GROUP;
MODEL Y1 Y2 Y3 Y4 Y5 = GROUP;
manova h=_all_ / printe printh;
然后你的output会有,Wilks' Lambda,Pillai's Trace,Hotelling-Lawley Trace等方法
的test结果..你自己选一个用就行了.
至于要不要用
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)