Comments on: Why practitioners discretize their continuous data

By: discrete data

discrete data — Tue, 30 Mar 2010 01:27:22 +0000

[...] setbacks. ... adopted DRT (Discrete Reportable Transcription) technology recognizes the need for ...10001036 : Why practitioners discretize their continuous dataWhy practitioners discretize their continuous data. Yihui asked this question yesterday. ... [...]

By: 谢益辉: 统计学博文导读：火箭队比赛与分类树、神经网络与降维 | 统计之都

谢益辉: 统计学博文导读：火箭队比赛与分类树、神经网络与降维 | 统计之都 — Sun, 15 Mar 2009 10:04:34 +0000

[...] 李晓煦老师的博客：非常专业，为数不多的会用LaTeX写上数学公式的博客，李老师对统计理论细节研究很认真，很有国外统计研究者的风范；博文如Why practitioners discretize their continuous data讲述了为什么大家喜欢将连续型数据离散化的原因之一。 [...]

By: lixiaoxu

lixiaoxu — Sun, 08 Mar 2009 09:56:27 +0000

Residuals and errors are different. The more intervals, squared-residuals decrease while squared-errors increase. So the black points, or discretization with max intervals, predict red population the worst.

Discretization fades micro information (most errors) while highlights macro information (usually non-linear). When LOESS is popular enough, discretization will be abandoned. Practitioners really need local smoothing to preview their concerned macro models.

By: Yihui

Yihui — Sun, 08 Mar 2009 07:26:08 +0000

The discretization here is essentially a kind of local smoothing techniques using a constant kernel function. Generally speaking, local modeling can effectively improve fitness (lower error sum of squares) but we have to carefully avoid overfitting. If you discretize x into more intervals, the fitting will be even better.

By: Keep on Fighting!

Keep on Fighting! — Fri, 06 Mar 2009 16:01:17 +0000

离散化：毁灭信息的有效手段...

如果你想掩盖数据，那么就把它们离散化吧！不知道为什么这么多人钟爱于将连续数据离散化，例如明明有年龄数据，在分析的时候非要分成老幼青壮这样的分类变量；明明有原始的计数数据...