What is NOT a recommended practice for handling missing data during EDA?

Question

BlackTom AI · Accepted Answer

In exploring the handling of missing data during EDA, it helps to evaluate each option on its own merits and limitations.
Option a: 'Removing affected rows if the percentage of missing data is low' — This can be a reasonable, pragmatic approach when missingness is minimal and unlikely to bias results. It preserves the integrity of analyses by excluding only a small subset. However, one must ensure that the removal doesn't disproportionately affect important groups or introduce bias, which means this method should be used judiciously rather than as an automatic rule.
Option b: 'Imputing using the mean or median' — Imputing with mean or median is a classic, simple imputation technique. It can reduce bias from missingness and keep dataset size intact, especially for numeric features. Yet, it ignores the potential relationships between features and the uncertainty about the true value, which can lead to underestimation of variance and distorted analyses if the data are not missing completely at random.
Option c: 'Leaving missing data unaddressed' — This stance is generally not recommended in EDA. Leaving gaps in the data can propagate uncertainty, hinder many analyses, and obscure patterns. Most analyses assume complete data matrices or, at minimum, acknowledged handling of missingness, so ignoring missing values is typically considered a poor practice.
Option d: 'Replacing with a constant value' — Imputing a constant (e.g., -1 or 'missing') is occasionally used, especially for categorical features or in certain modeling contexts, to explicitly mark missingness. However, it can severely distort distributions and relationships if not handled carefully, and it may mislead analyses that treat the substituted values as true observations.
Across these choices, the standout poor practice is leaving missing data unaddressed, since it yields incomplete information and undermines most downstream analyses. The other options offer viable strategies under appropriate conditions, with caveats about bias, variance, and the nature of the missingness.

Data Science Principles (202504-LLecture)

What is NOT a recommended practice for handling missing data during EDA?

查看解析

登录即可查看完整答案

类似问题

Which step is essential in handling missing data?

Creating a dummy variable to indicate a missing predictor value is a valid way to handle missing values

In our course, if I don't have any information about a specific Marin Factor I should assign a value of:

In a consumer society, many adults channel creativity into buying things

Economic stress and unpredictable times have resulted in a booming industry for self-help products

People born without creativity never can develop it

A product has a selling price of $20, a contribution margin ratio of 40% and fixed cost of $120,000. To make a profit of $30,000. The number of units that must be sold is: Type the number without $ and a comma. Eg: 20000

更多留学生实用工具

考试浏览器助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单