现在已经出了新版

the attachment image is not shown!

Bureau of Investigative Journalism https://www.thebureauinvestigates.com/ The Bureau of Investigative Journalism is an independent, not-for-profit organisation that holds power to account. Founded in 2010 by David and Elaine Potter, we tackle big subjects through deep reporting that uncovers the truth. We tell the stories that matter.

维基解密 wikileaks

big data literacy？

迈耶，应保持翻译一致性

https://baike.baidu.com/item/%E6%96%B0%E6%96%B0%E9%97%BB%E5%AD%A6/5733346 new journalism

collect survey data via questionaries and conduct cross tab analysis. Phlip Myers.

Wind of change How a team of Miami Herald reporters correlated hurricane wind speeds and insurance damage reports to win a Pulitzer prize JT Johnson Mon 3 Jul 2000 10.47 EDT Hurricane Andrew crashed down on the south Florida coast in August 1992. When it had passed, it left as its legacy the most expensive natural disaster in US history. Damage exceeded $20bn. More than 40 people were killed and approximately 2m Floridians were evacuated from their homes.

In the weeks that followed, Steve Doig and his colleagues at the Miami Herald sought to put the storm in perspective. Doig employed computerised mapping software - then unknown in newsrooms - to create maps correlating property tax assessment records and plots of the storm track.

Doig was able to show his readers that if Hurricane Andrew had made landfall just 20 miles farther north, it would have roared through the heart of Miami. Physical damage, and quite possibly the death count, could have more than tripled.

Advertisement

That alone was a good and imaginative piece of journalism, but Doig did not stop there. Using the same computing technology, he put together maps that matched geographic-specific wind speeds with building damage reports. He was intrigued to find there was more damage in areas that experienced lower wind speeds.

Having made the invisible visible thanks to the mapping and database software, Doig and other reporters followed their technological lead with the traditional journalist's tool: reporting. They hit the streets to seek explanations from public administrators, insurance companies and contractors as to how lower hurri cane winds could produce more damage than high winds.

What the reporters found was a litany of shoddy inspection by public officials. A few months later, the Miami Herald team was awarded the Pulitzer Prize for coverage that "not only helped readers cope with Hurricane Andrew's devastation but also showed how lax zoning, inspection and building codes had contributed to the destruction".

The digital tools that Doig was using are called geographical information systems or, more commonly, GIS. GIS represents the merger of cartography and digital databases, working together under the cover of today's desktop computer.

https://en.wikipedia.org/wiki/Bill_Dedman

网络科学对于社会网络分析方法，尤其是指数随机图模型及其变体介绍不足。尤其是Rsiena http://www.stats.ox.ac.uk/~snijders/siena/

如果是多分类的因变量，可以使用MNLogit模型

https://www.statsmodels.org/stable/generated/statsmodels.discrete.discrete_model.MNLogit.html

sklearn里面的数据表征形式

为什么使用交叉熵？p表示真实标记的分布，q则为训练后的模型的预测标记分布，交叉熵损失函数可以衡量p与q的相似性。

交叉熵作为损失函数还有一个好处是使用sigmoid函数在梯度下降时能避免均方误差损失函数学习速率降低的问题，因为学习速率可以被输出的误差所控制。

逻辑回归配合MSE损失函数时，采用梯度下降法进行学习时，会出现模型一开始训练时，学习速率非常慢的情况

解放 vs 建设！

妇代会现在还有吗？

其实也适用于其他社会科学领域。

这本书如同网络群体与市场（networks, crowds, and markets）一书，看上去简单却着实深刻，同时需要下功夫才能真正领会。

仅仅学习python等编程语言是不够的，更重要的是要在研究中真正使用它。经过大数据冲刷后的python使用经验更加宝贵。

重大问题是社会建构的，具体而言是社会当中的一部分人决定的，而非研究者自己决定的。为了论证一个问题是重要的问题，研究者需要确切地阐述问题（formulate the problem）。

The burden of knowledge!

这是一个重要问题。

我坚信大数据+因果推断是未来的研究趋势

不应被看作教科书，而更像是古今图书集成，或者说讲义。

这好像是爱因斯坦的话❀

个性化的信息推荐系统将会变革人类的信息消费模式。

产业驱动学术研究和教学

后来被改成了计算社会科学。

图片不显示呀！

这个功能在抓取腾讯较真数据的时候会使用到。

这一个是一级标题，需要修改。

add a website link