j****n
发帖数: 429 | 1 12-Week Summer Internship, Natural Language Processing and Text Mining
http://us.gsk.com/html/career/career-summer.html (search for requisition number: 73364)
GlaxoSmithKline (GSK) is a one of the world's largest pharmaceutical company
. We develop medicine and vaccines for both major disease areas including
asthma, cancer, virus and infections, mental health, diabetes and digestive
conditions. In addition, we are a leader in the important area of vaccines
and are developing new treatments for c... 阅读全帖 |
|
j*******1
发帖数: 266 | 2 按照现行FDA的规则，一个化合物只有通过安全性和有效性检验才能得到批准, 进入市
场。在经过三期临床试验后，至少有1/2 药物因效果不佳不能得到批准或根本没有申报。
临床试验无效意味什么呢？外行人就会认为药物临床试验没有效果，就是说它和安慰剂
的一样，一定是心理因素的结果，其实这个想法是很片面的，是对药物批准的过程不了
解。我以一个新药物的临床试验为例。
一个化合物要成为药物，临床试验三期是必不可少的。药物临床试验应该用双盲法，一
期实验是找一些志愿者，看看人们对药物的承受能力，不是两组对照，而二和三期中，
就要在两组中对照了。对照中的两组可有三种可能：
1） 要测试的药与 安慰剂（淀粉片）比较
2） 要测试的药与 另外一种已经上市的药物比较，目的是看看要测试的药与比另外
一种已经上市的药物效果好
3） 要测试的药加上一种已经上市的药物与 另外一种(两种)已经上市的药物比较，
目的是看看要测试的药能否在原来的药物的基础上有进一步的效果。
可以看出这三种情况，都只有一种是被测试的药物。如果临床试验无效，在第一和第三
种情况下，可以说要测试的药与安慰剂（淀粉片）一样；在第... 阅读全帖 |
|
j*******1
发帖数: 266 | 3 按照现行FDA的规则，一个化合物只有通过安全性和有效性检验才能得到批准, 进入市
场。在经过三期临床试验后，至少有1/2 药物因效果不佳不能得到批准或根本没有申报。
临床试验无效意味什么呢？外行人就会认为药物临床试验没有效果，就是说它和安慰剂
的一样，一定是心理因素的结果，其实这个想法是很片面的，是对药物批准的过程不了
解。我以一个新药物的临床试验为例。
一个化合物要成为药物，临床试验三期是必不可少的。药物临床试验应该用双盲法，一
期实验是找一些志愿者，看看人们对药物的承受能力，不是两组对照，而二和三期中，
就要在两组中对照了。对照中的两组可有三种可能：
1） 要测试的药与 安慰剂（淀粉片）比较
2） 要测试的药与 另外一种已经上市的药物比较，目的是看看要测试的药与比另外
一种已经上市的药物效果好
3） 要测试的药加上一种已经上市的药物与 另外一种(两种)已经上市的药物比较，
目的是看看要测试的药能否在原来的药物的基础上有进一步的效果。
可以看出这三种情况，都只有一种是被测试的药物。如果临床试验无效，在第一和第三
种情况下，可以说要测试的药与安慰剂（淀粉片）一样；在第... 阅读全帖 |
|
d****i
发帖数: 11 | 4 The orientation of each stick is an independent r.v., by
restricting all sticks to lie in a plane, the problem is
easy to solve for N=2. When N>=3, the distance between
endpoints becomes a function of 2 or more r.v.'s, then
how do you compute pdf of distance? |
|
S*********g
发帖数: 5298 | 5 http://www.physics.ucsb.edu/~jpierre/strings/dbranes.htm
____________________________________________________________
____________
SUPERSTRINGS! D-branes
Strings can have various kinds of boundary conditions. For
example closed strings have periodic boundary conditions
(the string comes back onto itself). Open strings can have
two different kinds of boundary conditions called Neumann
and Dirichlet boundary conditions. With Neumann boundary
conditions the endpoint is free to move about but no
mo |
|
jl
发帖数: 398 | 6 ☆─────────────────────────────────────☆
sonia (sonia) 于 (Wed Aug 3 15:06:37 2005) 提到:
google了一圈,却没看出这到底是指什么统计方法,用在什么情况下. 请各位略为介绍一下
可以么?或者提供一个网站或者文章阅读一下? 非常感谢你的帮助!
☆─────────────────────────────────────☆
tea (凤凰单枞) 于 (Wed Aug 3 18:06:32 2005) 提到:
I will only speak for the 'ad hoc analysis' as I understand it in
clinical trials.
In clinical trials, you analyze the data according to your analysis plan
(AP). Everything has to be pre-specified in your AP (endpoints,
statistical methods, e |
|
a********a
发帖数: 346 | 7 My data and sas code are as following.
我想对score做个histogram，在这个histogram上，我想让x轴上显示score，y轴上显示
relative frequency。但是用上面的命令，结果y轴上显示的是percent。不知道怎么能
让y轴显示relative frequency,而不是percent。 太感谢了。
data one;
input score @@;
datalines;
73 72 77 79 80 82 84 84 86 80
;
run;
proc univariate;
histogram score/ endpoints=64.5 to 95.5 by 5;
run; |
|
s*r
发帖数: 2757 | 8 you are looking for the thing mentioned by DaShagen
google "trend test proc glm contrast"
,
estimated
endpoints |
|
l******0
发帖数: 313 | 9 不是牛人，不过我觉得你这组数据是否有遗漏呢？time [6,8)的数据在哪里？另外，
time statement里面，我觉得用的应该都是endpoint,即我们所说的death time or on
study time
希望对你有帮助，如果有错误也请牛人帮忙指证：-） |
|
j*****e
发帖数: 182 | 10 Another question, are the endpoints of your interval random or fixed? |
|
s*******e
发帖数: 1385 | 11 endpoint是我record也就是检查有多少虫子死了，或者missing。有时候是两天，有时
候是三天。应该算是random的吧。 |
|
D******n
发帖数: 2836 | 12 按逻辑来说，要先确定4，what is the endpoint, what is the surrogate variable? |
|
d*******1
发帖数: 854 | 13 数据是这样的（简单的例子）
gene treatment value
A drug 12
A drug 13
A vehicle 14
A vehicle 20
B drug 456
B drug 400
B vehicle 200
B vehicle 100
就是用LM()对drug和vehicle做t test, 但是有很多gene, 我知道可以用loop, 但是肯
定很慢， 用tapply可能好些， 那位给提个思路？
谢谢了。 |
|
D******n
发帖数: 2836 | 14 lm(value~gene+treatment,data) |
|
p***r
发帖数: 920 | 15 by(data,gene,function(x) lm(value~treatment)) |
|
d*******1
发帖数: 854 | 16 gene不是用在model里的， 是作为by variable |
|
d*******1
发帖数: 854 | 17 这这里这个by可以用lapply什么的代替呢， 两者之间有什么区别？ |
|
p***r
发帖数: 920 | 18 good point, i dont know, any one have 2cent? |
|
D******n
发帖数: 2836 | 19 lapply(split(data,data$gene),function(x) lm(value~treatment,x)) |
|
p***r
发帖数: 920 | 20 so lapply do apply the function to a list, but you first should have a
list, otherwise, it will just take each column as a list? |
|
D******n
发帖数: 2836 | 21 it applies fun to the element of a list. so if u use it on data.frame, it ap
plies fun to each column. |
|
d*******1
发帖数: 854 | 22 那看来在我的情况下用by还是比较简单， 能不能总结下什么情况下用lapply呢？（还
有apply, sapply等等） |
|
d*******1
发帖数: 854 | 23 我还想就这个算法的performance请教一下大家， 我这个数据有大概6万个gene (by
variable), 在每个基因下是一个6对6的t test,大家觉得应该用多长时间搞定。 我在
一个linux server (8 cores, 32 GB memory), 耗时3-4小时还没有搞定。会是什么问
题呢 |
|
l****a
发帖数: 352 | 24 many to many merge is the most common log statement that you need to check
in the log window
other things that you need to notice include error,warning and uninitiate
MOST industry use noncopartmental method for early phase study and there is
almost no stat there.Data/model are more complex for phase iii/iv studies.
You need to pay attention to the study procedure by visits, the endpoints
defination (i.e.,LOCF) and the data/model sepec.
Also as a sas programmer, I think the key point is to build |
|
e****t
发帖数: 766 | 25 只是在课上学过简单的计算和定义。
有没有系统的书介绍一下不同distribution endpoint 的sample size 的calculation
主要是用在clinical trial上面。
谢谢了！ |
|
s*r
发帖数: 2757 | 26 thx
按地域来pool是仅适用于international trial,还是在美国国内的trial也适用?
ENDPOINT
SIZE |
|
p********a
发帖数: 5352 | 27 唉，我做HEALTHCARE CLAIMS也做腻歪了。天天都是那么些东西，我闭着眼睛也搞得出
来。主要的程序全被我MACRO化了，写什么程序都是换几个参数就行了。
我也想转CLINICAL TRIALS，也努力过，还专门上过课。不过一想到从年薪6万的JUNIOR
做起就受不了，还得养家啊。
ENDPOINT
SIZE |
|
w********o
发帖数: 1621 | 28 statistician还真是越老越吃香。做过的trial越多，什么药都知道用什么做endpoint
，怎么study design，怎么统计分析。这样的人，我是佩服得紧。
软工都有这问题，老了如果不是升上去做manager，最后八成要被年轻的淘汰。在编程
上，那效率真是差一大节。对sas programmer来说，senior的价值应该是懂很多工业界
的standard。如果到了senior还在编程，那劣势真的是挺明显的。至少我自己的
project lead programmer一定找效率高的，耗不起那时间。
我倒不是歧视senior，我只是在说效率问题而已。在哪儿做，效率都很重要。至少
statistician肯定愿意跟干活又快又好的sas programmer合作。 |
|
d********t
发帖数: 837 | 29 There are many more , the most popular choice for continuous endpoints
is based on fully specified likelihood and covariance modeling,
possibly with random effects. GEE is often used for categorical
data,but almost never seen with continuous data in practice (maybe
gamma family?). As the usual interest in a clinical trial is for the
treatment effect at a certain timepoint, longitudinal models are mostly
used to account for missing data over time (the only exception I've
seen is to estimate coeff... 阅读全帖 |
|
e****t
发帖数: 766 | 30 During work, all the binary endpoint, we use fisher. i am not sure whether
this happens to other pharms. |
|
h******m
发帖数: 8 | 31 Get the survival probability first, cumulative incidence or mortality is 1-
survival.
by the way, if you have competing risk in the endpoint, use %cuminc macro in
SAS. The macro can only give you the mortality curves but not point
estimates . you need to use R to do that at this time. Hope it helps. |
|
b***8
发帖数: 60 | 32 In biomarker studies, the ultimate goal is to screen dozens, or
hundreds, or potentially hundreds of thousands of potential predictors
using explanatory methods to understand the existing data and then
ultimately identify a factor or set of factors that could be
used in the prediction of new observations with high accuracy. While
the concepts of "association testing" and "classification" are not
unrelated in this scenario, they are inherently separate objectives
requiring the implementation of d... 阅读全帖 |
|
n*****3
发帖数: 1584 | 33 wait longer or try other endpoints. |
|
z**********i
发帖数: 12276 | 34 Good morning!
We have a new part-time, telecommute Biostat opening with a client based out
of PA, and I am reaching out to see if you know of anyone currently looking
for a new opportunity. I have provided the job details below for your
review and politely ask that you forward this out to any friends or
colleagues that may be looking. While the client allowing this position to
work remotely, they do require that this candidate travel onsite once a
month for meetings, updates, etc. If you or anyo... 阅读全帖 |
|
h*********1
发帖数: 37 | 35 I received the opening from a recruiter. If you are interested, please
contact the recruiter directly or me (t******[email protected]), I can help to
forward the resume.
Title: Clinical Biostatistician
Company: Mid Size Pharma
Location: Home Based (must reside in Eastern PA, DE, NJ, NY)
Hours: PT 10-30 hours per week
Length: initial 6 month contract
Position Summary
The Statistical Consultant position is for a skilled scientist with
training and significant experience in statistics, drug development... 阅读全帖 |
|
s***1
发帖数: 343 | 36 Our company is currently looking for an in-house Clinical Data Manager.
Please see below. If interested, please send your resume to stellaszh@gmail.
com. I'll represent it to our HR. Thanks.
Job Summary
We are seeking a Clinical Data Manager (CDM) who will serve as the Lead
Clinical Data Manager across studies and liaison between Clinical Research,
Data Management, and Statistical Analyses. This position works with minimal
supervision in consort with the in-house management.
Essential Duties an... 阅读全帖 |
|
s*****e
发帖数: 3954 | 37 看到一个CRO的statistical analyst job description里有条responsibility是
writing SAS programs to validate variable within datasets, key endpoints and
denominators within tables and figures.
一般都做什么validation？ |
|
xi
发帖数: 1678 | 38 INTERVALS=
specifies interval endpoints for life-table estimates
is this one? |
|
l*******o
发帖数: 71 | 39 想今秋转行学统计，请有经验的同胞们给点建议和鼓励给我指条明路。现在的我很
迷茫不知道自己的选择是不是正确的。最近看过很多帖子说统计小硕找工作不容易什么
的，弄得我一身冷汗。先做一下自我介绍本人31了，2009年毕业于国内一普通大学2E专
业。来美国5年了一直做全职，现在宝宝也快2岁了。我和老公的绿卡正在排期，大概是
两年后能排到，所以想现在学点东西到时候找个工作。我现在的想法就是只要毕业能找
到工作就好。真怕花了老公大把银子最后还是家庭主妇。哦对了忘交代了，我的英语水
平是一般。教育宝宝都是全中文的。我的问题主要有2个方面。
1.学统计从事哪个方向的就业机会多啊？我看有学生物统计去药厂的，有在银行和保险
公司的。看有人说药厂的工作经验到换工作的时候是很难转的。那什么方向的工作经验
是越久越光明呢？
2学校的curriculum上有好多课程，有经验的兄弟姐妹们能帮我看看哪些课程的实用性
强么？里边有的应该是phd的课程，不知道我能不能选呢。我把一些我觉的不用选的和
不能选的课程删掉了。小硕就是10门课。
courses list：
8001. Probability ... 阅读全帖 |
|
b*********n
发帖数: 182 | 40 假如你手上有一个有关clinical trial的大型数据库，主要信息包括各个clinical
trial的基本信息，如 product name, company name, primary endpoint, randomized
or not, sample size, blindness, etc. 你会做些什么分析，最想用这个数据库回答
什么问题？ |
|
l******1
发帖数: 12 | 41 If you are interested, please send your resume to william.liu@
techdataservice.com.
Job Category: Clinical and Scientific
Job Code:
Job Title: Biostatistician III
Location: East Hanove, New Jersey
Duties:
Designing clinical studies. Writing and reviewing statistical sections and
other relevant parts of study synopses and protocols. Support study start up
activities.
Develop RAP, program TLFs, preparation for database lock, dry run and CSR
activities; review CRF, DB spec, VAP; su... 阅读全帖 |
|
j*****e
发帖数: 182 | 42 Hot topics in clinical trial:
equivalence/non-inferiority tests,
multiple testing procedures,
design and power analyses associated with the above tests,
dose-response studies,
analysis of safety endpoints,
clinical trials on vaccines (biosimilar studies),
etc. |
|
s**e
发帖数: 294 | 43 Strategically，你的trial要给label增加什么statement？在这个基础上，几个
treatment group？几个endpoints？有没有hierachy？alpha怎么分配最efficient？要
不要prespecify subgroup？要不要interim？要不要adaptive？missing value怎么处
理？当然根据你trial的phase还有侧重。 |
|
d********t
发帖数: 837 | 44 A1c , fasting plasma glucose . 用longitudinal model. |
|
B******y
发帖数: 9065 | 45 你就一组病人，连个参照组都没有，两个时间段用paired t-test最简单了。什么mixed
model和repeated measure ANOVA纯粹是扯淡。。。建议你先画个图，指标随时间的变
化一目了然，你心里就有数了。干脆取Change from Baseline做endpoint，用药后的每
个时间点都做个t test，报告里面然后放些p就好看一点。。。 |
|
s**********3
发帖数: 9 | 46 Hi! If any one is interested in the following opportunity of a Senior SAS
programmer in a healthcare company, please submit your resume to the
recruiter:
Stephen Anglin ([email protected]
/* */)
The recruiter is a very nice guy!
Job description:
Contract SAS programmer to support data validation in Data Management and
Biostatistics functions. Perform QA activities for the Biostatistics area
in support of the design and interpretation of clinical research/trial data.
RESPONSIBILITIES:
... 阅读全帖 |
|
B******y
发帖数: 9065 | 47 ADaM
1. Which variable metadata field is NOT a part of a Version 2.1 ADaM-
compliant analysis dataset?
A. Dataset Name
B. Variable Label
C. Source/Derivation
D. Display Format
E. Codelist/Controlled Term
F. Variable Type
G. Informat
2. According to ADaM, the structure for most datasets should be vertical (i.
e., tall and thin).
A. TRUE
B. FALSE
3. In ADaM as in the SDTM, there is no day 0 for relative day variables
whose name ends in DY.
A. TRUE
B. FALSE
4. In ADaM-compliant datasets the display... 阅读全帖 |
|
i***m
发帖数: 148 | 48 我觉得是要对制药的整个流程都有参与，其实很多时候这些东西与统计方法无关，
比如在一个新的疾病领域，你怎么identify一个可以拿去跟FDA申请的endpoint？这是
从preclinical就要开始考虑，
stats就要开始invovle的。
长久来看还是这些对制药流程的领悟更重要
很多时候并不是技术，而是一些“business” sense吧 |
|
B******y
发帖数: 9065 | 49 谈不上什么业内人士，聊一些个人看法吧，就当抛砖引玉，大家可以讨论。
首先，P值只是个参考数据，本身不是决定性因素。当然，一个trial最终的endpoint要
是P值大于0.05肯定是没戏啦（假定我们讨论的不是non-inferiority study），但即使
小于0.05也说明不了什么问题。一份Protocol的设计本身就有很多的限制条件，一般都
是针对某些特定的病人组群的，从clinical pathology就有很多假定。即使到了统计部
分，模型的运用和数据的收集也都是基于一些统计理论假定。也就是说，当P值最终被
计算出来，都是经过了一系列的临床的和统计上的理论假设。万一有一点的假设不成立
，结论就会被很脆弱的攻击。而且Clinical Trials的结果应该具备可复制性。要是设
计相差不大的3个trial都做了，2个主P值小于0.05，而另一个大于0.05，如何解释？这
种情况在Phase 3阶段太常见不过了。当然你可以通过ISE来整合数据，这也是FDA/EMA
都要求。也就是说，P值只能反应出某个临床实验在某个特定情况下实现的，是否能通
用到更广泛的情况需要监管部门深思熟虑... 阅读全帖 |
|
e****t
发帖数: 766 | 50 是不是primary endpoint 是 os ， 可以将death patients 当作censoring 认为 如果
不died 会fu 大于death 时间 所以应用survival technique 去估计median followup
什么时候用median of the observed value 合适？ |
|
