illustrates the use of concepts related to the research procedure; and proposes measures to achieve trustworthiness (credibility, dependability and transferability) throughout the steps of the research procedure.

Interpretation in qualitative content analysis is discussed in light of Watzlawick et al.’s [Pragmatics of Human Communication. A Study of Interactional Patterns, Pathologies and Paradoxes.

Interpretation in qualitative content analysis is discussed in light of Watzlawick et al.’s [Pragmatics of Human Communication. A Study of Interactional Patterns, Pathologies and Paradoxes.

Initially content analysis dealt with ‘the objective, systematic and quantitative description of the manifest content of communication’ (Berelson, 1952, p. 18) but, over time, it has expanded to also include interpretations of latent content.

Many authors, from a variety of research traditions, have addressed content analysis (for example, Berelson, 1952; Krippendorff, 1980; Findahl and Höijer, 1981; Woods and Catanzaro, 1988; Downe-Wamboldt, 1992; Burnard, 1991 and Burnard, 1996; Polit and Hungler, 1999).

The first descriptions date from the 1950s and are predominately quantitative. Currently, two principal uses of content analysis are evident.

The first descriptions date from the 1950s and are predominately quantitative. Currently, two principal uses of content analysis are evident.

One is a quantitative approach often used in, for example, media research, and the other is a qualitative approach often used in, for example, nursing research and education.

与学术无关的话: nice.

与学术无关的话:  just chatting.

What sort of an act do you do? I bend over backwards and pick up a handkerchief with my teeth.

Linear mixed-effects models (LMMs) are an important class of statistical models
that can be used to analyze correlated data. Such data include clustered
observations, repeated measurements, longitudinal measurements, multivariate
observations, etc.

The aim of our book is to help readers in fitting LMMs using R software. R
(www.r-project.org) is a language and an environment aimed at facilitating
implementation of statistical methodology and graphics. It is an open-source
software, which can be freely downloaded and used under the GNU General
Public License. In particular, users can define and share their own functions, which
implement various methods and extend the functionality of R. This feature makes R
a very useful platform for propagating the knowledge and use of statistical methods.

We believe that, by describing selected tools available in R for fitting LMMs,
we can promote the broader application of the models. To help readers less familiar
with this class of linear models (LMs), we include in our book a description of the
most important theoretical concepts and features of LMMs. Moreover, we present
examples of applications of the models to real-life datasets from various areas to
illustrate the main features of both theory and software.

There are many packages in R, which contain functions that allow fitting various
forms of LMMs. The list includes, but is not limited to, packages amer,
arm, gamm, gamm4, GLMMarp, glmmAK, glmmBUGS, heavy, HGLMMM,
lme4.0, lmec, lmm, longRPart, MASS, MCMCglmm, nlme, PSM, and pedigreemm.
On the one hand, it would seem that the list is rich enough to allow for

On the other hand, the number of available packages
leads to difficulty in evaluating their relative merits and making the most suitable
choice.

In the coming big data era, statistics and machine learning are becoming
indispensable tools for data mining. Depending on the type of data analysis,
machine learning methods are categorized into three groups:

Supervised learning: Given input-output paired data, the objective
of supervised learning is to analyze the input-output relation behind the
data. Typical tasks of supervised learning include regression (predicting
the real value), classification (predicting the category), and ranking
(predicting the order). Supervised learning is the most common data
analysis and has been extensively studied in the statistics community
for long time. A recent trend of supervised learning research in the machine
learning community is to utilize side information in addition to the
input-output paired data to further improve the prediction accuracy. For
example, semi-supervised learning utilizes additional input-only data,
transfer learning borrows data from other similar learning tasks, and
multi-task learning solves multiple related learning tasks simultaneously.

Unsupervised learning: Given input-only data, the objective of unsupervised
learning is to find something useful in the data. Due to this
ambiguous definition, unsupervised learning research tends to be more
ad hoc than supervised learning. Nevertheless, unsupervised learning is
regarded as one of the most important tools in data mining because
of its automatic and inexpensive nature. Typical tasks of unsupervised
learning include clustering (grouping the data based on their similarity),
density estimation (estimating the probability distribution behind the
data), anomaly detection (removing outliers from the data), data visual-
ization (reducing the dimensionality of the data to 1–3 dimensions), and
blind source separation (extracting the original source signals from their
mixtures). Also, unsupervised learning methods are sometimes used as
data pre-processing tools in supervised learning.

Reinforcement learning: Supervised learning is a sound approach,
but collecting input-output paired data is often too expensive. Unsupervised
learning is inexpensive to perform, but it tends to be ad hoc.
Reinforcement learning is placed between supervised learning and unsupervised
learning — no explicit supervision (output data) is provided,
but we still want to learn the input-output relation behind the data.

Instead of output data, reinforcement learning utilizes rewards, which
evaluate the validity of predicted outputs. Giving implicit supervision
such as rewards is usually much easier and less costly than giving explicit
supervision, and therefore reinforcement learning can be a vital
approach in modern data analysis. Various supervised and unsupervised
learning techniques are also utilized in the framework of reinforcement
learning.

This book is devoted to introducing fundamental concepts and practical
algorithms of statistical reinforcement learning from the modern machine
learning viewpoint. Various illustrative examples, mainly in robotics, are also
provided to help understand the intuition and usefulness of reinforcement
learning techniques. Target readers are graduate-level students in computer
science and applied statistics as well as researchers and engineers in related
fields. Basic knowledge of probability and statistics, linear algebra, and elementary
calculus is assumed.

[一种声音]胡戈：带儿子看病的全过程

我刚刚一岁零几天的孩子咳嗽，外加低烧37.7（对一岁婴儿来说还不算发烧），人还是挺活泼的，与平时相比没什么不同。老婆和岳母带他去某家大型儿童医院。我因为忙没有去。

本来应该挂呼吸科，但因为是双休日，呼吸科不上班，于是挂了内科普通门诊……做了血常规，显示某些项目偏低，诊断为“上感”。开了四种口服药和雾化。



老婆把单子拍给我看，我一看就心想：糟了！

第一种是抗生素。抗生素对病毒无效这是我知道的。

第二种施保利通片。我还以为是什么西药，一查，我了个去，是草药！德国草药，作为中药进口的。

进口药品　德国 Schaper & Brümmer GmbH & Co. KG
英文名称：Esberitox N Tablets
注册证号：Z20090002，2009-03-30
药品特性：中药，每片重0.3g


这药特别神，教科书上没有，用药指南里没有，却被国内的儿科作为常用药来用，说是能提高免疫力，特别喜欢开。在德国是非处方草药，四岁以下儿童禁用，到了中国变成处方药，改配方改说明书，没有了年龄限制。非儿科医生大都不认识这玩艺。知乎上有几个医生在我对此药进行声讨时试图为它进行辩护，拿着说明书说没问题，因为说明书上写了可以用于婴儿，在别人指出问题后依然不屈不挠试图为此药正名。

这药70元一盒，是开出的所有药里最贵的。

据后来网友说，此药是药企行为。

第三种，清开灵颗粒，中成药。

第四种，盐酸伪麻黄碱 和 氢溴酸右美沙芬，也就是艾畅。右美沙芬是一种中枢止咳药，FDA禁止2岁以下儿童使用。

我查完资料后我就崩溃了。老婆要听医生的，我跟她吵起来。她说她有一个手机 app 可以打电话问别的医生，我当然不信，但是好歹就让她打吧。


一个其他医院的儿科医生跟我们通上了话，详细了解了情况，看了医院的检查报告，说了挺久的，最后开始说这些药。直接把抗生素以外的三种药全部否定了。中成药，没有证据显示有效。艾畅，不要吃。施保利通片是德国的草药，不要吃。最后仅剩下抗生素，他说他很为难，因为按照美国来说是不会开的，但是国内都会开。上感只有极小的可能性是细菌性感染。他让我们自己决定吃不吃。当然，我老婆还是决定吃，真不让吃的话，我的家庭就要破裂了。

于是就开始做雾化治疗和服用抗生素。

吃抗生素后第二天，孩子开始呕吐、腹泻。我岳母就火了，说是我不让吃药造成，要把所有的药都给孩子吃。我一查，那个抗生素有呕吐、腹泻的副作用，于是怀疑是抗生素造成。但是我说话没用。

然后全家出动去复诊，这次说什么也要找个专家门诊。还是这个医院，很幸运地挂上了内科专家门诊。

体温36.5，血检指标基本正常，X光片显示肺部纹理增多（报告书上的意见是“支气管炎改变”），喉咙发炎……我向医生询问呕吐会不会是抗生素造成？她没有明确回答，或者声音太小我没听清。最后开药，说要挂水。

挂水这俩字一出来，我就傻眼了。然后说施保利通片，要吃！我当时就崩溃了。

还好她说清开灵可以不吃。这是唯一我感到欣慰的地方。

接下来在医院，我和岳母展开了抢婴儿大战。岳母坚决要挂水，我瞅准机会想把孩子抢跑，没成功，反而被岳母一阵大骂，还要老婆跟我离婚。眼看家庭就要破裂。


期间还用那个 app 联系了另一个医生，那个医生听说我们是在这个医院看的，就说这个医院没问题。我又崩溃。看来这个 app 也不是很靠谱，关键还是要看医生。最后好不容易联系上当初那个医生，他说他没办法给我们拿主意，因为有1%-5%的可能性是细菌感染。

最后岳母妥协，先不挂水，第二天去我们常去的一家私立医院（不是莆田系那种，千万不要去莆田系！）做最后的诊断和决定，医生说啥就是啥。这个医院是我们经常带孩子去做检查和打疫苗的，老婆和我都在那里看过病，很信任，但是比较远，收费也特别高（挂号费800起），所以这次感冒觉得是小病就没有去。

半夜里，孩子又是咳嗽又是呕吐。岳母又是一阵骂。

第二天，终于见到了我们信任的医生。她一看我们带去的那些药就皱眉。做完详细检查后，她解释了为什么这是病毒性而不会是细菌性感染，说孩子得的是由呼吸道合胞病毒感染引起的毛细支气管炎，在北半球这个纬度这个季节很常见，抗生素无效，除了雾化以外，那些药都属于过度治疗。腹泻确实是那个抗生素的副作用，清开灵会更厉害。咳嗽是支气管痉挛引起（我没听清，孩子在叫，我抱他出去了几次），也就是由“喘”引起，现在要控制“喘”，否则可能会演变成哮喘。开了三种控制“喘”的药，价格都不高。

这下岳母没话说了，我也满意了。家庭免于破裂。


后面我的想法：

1，公立医院的问题根源在于国家限制诊疗费，逼着医院以药养医，引发了一系列问题……

2，当然，如果不限制诊疗费，穷人看不起病怎么办？这我不知道，我只知道是谁造成了穷人那么穷的，就应该由它来解决这个问题。

3，我自己在公立医院的看病体验还是可以的，虽然人多拥挤，但是费用实在是低，而且基本不会乱开药。只是这一次我实在理解不了。我知道医生不容易，但真正到了我头上，而且是宝贝儿子，忍不了。

4，有理想有抱负的医生，可以考虑一下跳出体制。通过这次事件，我对体制外医疗又多了一份好感。

来源：胡戈

Identification as a randomised trial in the title

Structured summary of trial design, methods, results, and conclusions (for specific guidance see CONSORT for abstracts)

Scientific background and explanation of rationale

Specific objectives or hypotheses

Description of trial design (such as parallel, factorial) including allocation ratio

Logistic Regression in R – Part One Sep 2015
更新于 Wed Sep 02, 2015 17:54 由 atmathew 通过 R-bloggers
(This article was first published on Mathew Analytics » R, and kindly contributed to R-bloggers)
Please note that an earlier version of this post had to be retracted because it contained some content which was generated at work. I have since chosen to rewrite the document in a series of posts. Please recognize that this may take some time. Apologies for any inconvenience.


Logistic regression is used to analyze the relationship between a dichotomous dependent variable and one or more categorical or continuous independent variables. It specifies the likelihood of the response variable as a function of various predictors. The model expressed as log(odds) = \beta_0 + \beta_1*x_1 + ... + \beta_n*x_n, where \beta refers to the parameters and x_i represents the independent variables. The log(odds), or log of the odds ratio, is defined as ln[\frac{p}{1-p}]. It expresses the natural logarithm of the ratio between the probability that an event will occur, p(Y=1), to the probability that an event will not occur, p(Y=0).

The models estimates, \beta, express the relationship between the independent and dependent variable on a log-odds scale. A coefficient of 0.020 would indicate that a one unit increase in \beta_i is associated with a log-odds increase in the occurce of Y by 0.020. To get a clearer understanding of the constant effect of a predictor on the likelihood that an outcome will occur, odds-ratios can be calculated. This can be expressed as odds(Y) = \exp(\beta_0 + \beta_1*x_1 + ... + \beta_n*x_n), which is the exponentiate of the model. Alongside the odd-ratio, it’s often worth calculating predicted probabilities of Y at specific values of key predictors. This is done through p = \frac{1}{1 + \exp^{-z}} where z refers to the log(odds) regression equation.

Using the GermanCredit dataset in the Caret package, we will construct a logistic regression model to estimate the likelihood of a consumer being a good loan applicant based on a number of predictor variables.

library(caret)
data(GermanCredit)

Train <- createDataPartition(GermanCredit$Class, p=0.6, list=FALSE)
training <- GermanCredit[ Train, ]
testing <- GermanCredit[ -Train, ]

mod_fit_one <- glm(Class ~ Age + ForeignWorker + Property.RealEstate + Housing.Own +
CreditHistory.Critical, data=training, family="binomial")

summary(mod_fit_one) # estimates
exp(coef(mod_fit$finalModel)) # odds ratios
predict(mod_fit_one, newdata=testing, type="response") # predicted probabilities
Great, we’re all done, right? Not just yet. There are some critical questions that still remain. Is the model any good? How well does the model fit the data? Which predictors are most important? Are the predictions accurate? In the next few posts, I’ll provide an overview of how to evaluate logistic regression models in R.