登录查看更多内容

What statistical misconceptions do you need to avoid in research management?

由人工智能和领英社区提供技术支持

As a research manager, you need to ensure that your projects are based on sound statistical principles and practices. However, statistics can be tricky and confusing, and there are many common misconceptions that can lead to erroneous or misleading conclusions. In this article, we will discuss some of the most frequent statistical pitfalls that you should avoid in research management, and how to overcome them with proper planning, analysis, and communication.

此文章中的业界达人

由社区从 6 条内容中精选。了解更多

Prof. (Dr.) Balvir S. Tomar

International President - World Health Summit | Founder & Chancellor - NIMS University
Duy Pham

Communication Scientist and Product Analyst

1 Correlation and causation

One of the most basic and yet most misunderstood concepts in statistics is the difference between correlation and causation. Correlation means that two variables are related in some way, but it does not imply that one causes the other. Causation means that there is a direct or indirect causal link between two variables, and that changing one affects the other. To establish causation, you need to control for other factors that might influence the relationship, and use appropriate experimental designs and methods. Do not assume that correlation implies causation, or vice versa, without sufficient evidence and reasoning.

添加您的观点

Prof. (Dr.) Balvir S. Tomar

International President - World Health Summit | Founder & Chancellor - NIMS University
举报内容
Always consider the possibility of confounding variables that might affect the observed correlation. A confounding variable is an external factor that can create a false impression of a causal relationship between the variables being studied. For example, ice cream sales and drowning incidents may be correlated, but the confounding variable here could be the weather, as both tend to increase during the summer. Proper statistical analysis and experimental controls are necessary to identify and mitigate the influence of such confounding variables, ensuring more accurate and reliable conclusions.

已翻译

赞
Keith Wong, FRM, PhD

Specialised in Data Analytics and Quantitative Modelling
(已编辑)
举报内容
My favorite textbook example is the positive correlation between the length of one's shoelaces and one's IQ. You won't try to increase your IQ by replacing for longer shoelaces - the latter is not relevant to, or has no causation with, your IQ at all.

已翻译

赞

2 Sampling and generalization

Another common source of error in statistics is sampling and generalization. Sampling refers to the process of selecting a subset of the population to study, and generalization refers to the process of applying the findings from the sample to the whole population. To ensure valid sampling and generalization, you need to consider the size, representativeness, and variability of your sample, and the confidence level and margin of error of your estimates. Do not use a sample that is too small, biased, or heterogeneous, or make claims that are too broad, vague, or unsupported by your data.

添加您的观点

Rachel Lopes

Geologist | Commercial Analyst | CBSP | CRM | SPE | AAPG Youg Professionals
举报内容
In the context of predictive models, models that seek to predict a result based on a set of training data may be vulnerable to errors due to the variance of the values observed in the training and test sets, resulting in an excellent evaluation for a specific case, but presenting unsatisfactory performance on different data. This phenomenon is known as overfitting. To avoid this situation and maintain the generalization capacity of the model, a technique called cross-validation is used. This approach allows you to train and test the model with all available data, increasing robustness and reducing variance, a measure of dispersion that verifies the distance between arithmetic mean values.

已翻译

赞
Prof. (Dr.) Balvir S. Tomar

International President - World Health Summit | Founder & Chancellor - NIMS University
举报内容
Be mindful of the sampling method you use. Random sampling is generally preferred as it reduces selection bias and increases the likelihood that your sample is representative of the population. Stratified sampling, where the population is divided into subgroups and samples are taken from each, can help ensure diversity and reduce sampling error. Avoid convenience sampling, where participants are chosen based on ease of access, as this can lead to significant bias. Ensure your generalizations are grounded in solid statistical analysis and clearly communicate the limitations and scope of your findings to avoid misleading conclusions.

已翻译

赞

3 P-values and significance

A third frequent mistake in statistics is misinterpreting p-values and significance. A p-value is the probability of obtaining a result as extreme as or more extreme than the observed one, assuming that the null hypothesis is true. Significance means that the p-value is below a certain threshold, usually 0.05, indicating that the result is unlikely to occur by chance. However, a low p-value does not mean that the result is important, meaningful, or true, nor does a high p-value mean that the result is irrelevant, trivial, or false. A p-value depends on many factors, such as the sample size, the effect size, the test statistic, and the assumptions of the test. Do not rely solely on p-values and significance to draw conclusions, but also consider the practical and theoretical implications of your results.

添加您的观点

Prof. (Dr.) Balvir S. Tomar

International President - World Health Summit | Founder & Chancellor - NIMS University
举报内容
It's important to understand that p-values do not measure the probability that the null hypothesis is true or the probability that the observed data occurred by chance alone. They also do not provide a measure of the size or importance of an effect. It's crucial to complement p-value analysis with confidence intervals, which offer a range of values within which the true effect likely lies, and effect size measures, which provide a sense of the magnitude of the observed phenomenon. Always contextualize p-values within the broader research framework, including the study design, data quality, and real-world relevance, to avoid overemphasizing statistical significance at the expense of practical significance.

已翻译

赞

4 Confounding and mediation

A fourth common problem in statistics is confounding and mediation. Confounding occurs when a third variable affects both the independent and dependent variables, creating a spurious or distorted relationship between them. Mediation occurs when a third variable explains or accounts for the relationship between the independent and dependent variables, acting as a mechanism or a pathway. To identify and control for confounding and mediation, you need to use appropriate statistical techniques, such as stratification, adjustment, matching, or path analysis. Do not ignore or overlook the potential effects of confounding and mediation on your results, or confuse them with each other.

添加您的观点

5 Reporting and communication

A fifth common challenge in statistics is reporting and communication. Reporting refers to the process of presenting and summarizing your statistical results, and communication refers to the process of explaining and interpreting your results for your audience. To ensure effective reporting and communication, you need to use clear, concise, and accurate language, and avoid jargon, ambiguity, and exaggeration. You also need to use appropriate visual aids, such as tables, charts, and graphs, and follow the ethical and professional standards of your field. Do not report or communicate your results in a way that is misleading, confusing, or incomplete, or that violates the principles of honesty, transparency, and integrity.

Statistics is a powerful and essential tool for research management, but it also requires careful and critical thinking. By avoiding these common statistical misconceptions, you can improve the quality and credibility of your research projects, and make better decisions based on data and evidence.

添加您的观点

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?

添加您的观点

Duy Pham

Communication Scientist and Product Analyst
(已编辑)
举报内容
In my work as a statistical consultant, one of the biggest challenges is explaining to clients why there needs to be rationale for their research questions. Many clients come to me with a dataset and expect to find any significant results without a priori hypothesis development. While I only make recommendations on research design and statistical analyses, I do try to orient the clients back to thinking about their research questions and rationale for their research.

已翻译

赞

Research Management

+ 关注

给文章评分

我们借助人工智能创建了此文章。您认为这篇文章怎么样？

很棒不太好

举报此文章

查看全部

What statistical misconceptions do you need to avoid in research management?

1

2

3

4

5

6

1 Correlation and causation

2 Sampling and generalization

3 P-values and significance

4 Confounding and mediation

5 Reporting and communication

6 Here’s what else to consider

Research Management

给文章评分

感谢您的反馈

更多Research Management相关文章

更多相关阅读内容