# 18 Oct Should medical students learn statistics?

Statistics are often considered complex and far from daily clinical practice. Legitimately, as a doctor or student, one may wonder whether it is useful to learn biostatistics and to perform statistical analyses by oneself.

## Learn statistics?

It is quite possible to learn the basics of statistics without having a solid knowledge of mathematics. The important thing is to understand the fundamental concepts of the definition of null and alternative hypotheses, p-value, confidence intervals, and to be able to correctly interpret the results of the analyses. In short, the theoretical side of statistics can be approached in a simple and intuitive way.

## Perform the statistical analyses by yourself?

The practical side of statistics is provided by the conduct of statistical analyses. Statisticians are highly solicited people, and cannot respond to all requests for statistical analysis in a timely manner. And many back and forth are necessary to achieve a result. It is therefore common for the analysis to take several months. To speed up the process, it may be tempting to do your statistical analyses alone.

For example, it is possible to memorize the tests adapted to a particular situation (example: Student T test to compare averages between 2 groups, Chi2 test to compare proportions, etc), or simply refer to tables available on the internet. However, it must be understood that carrying out statistical analyses requires a high degree of **rigour**. It is common to obtain a false result if good practices are not followed. These include the verification of conditions;, that can easily be forgotten to check when not performing statistical analysis on a daily basis.

**In many disciplines, the theory learned requires to be practiced regularly, otherwise it will be forgotten.** This is even more true for statistical analyses, which are highly technical and require learning to work with complex software with their own programming language.

## Which software to learn? R, Python, SAS, Stata?

If you are planning a career in medical research, it seems important to have a solid theoretical foundation in statistics. However, be realistic: unless you give up clinical practice and focus on full-time research, do not plan to conduct your statistical analyses occasionally with professional statistical software, such as R, SAS or Stata. This is an inefficient approach: the investment of time is tremendous and wasted a few months later. This is without mentioning errors, either of procedure or code, which will not be detected.

I would therefore not encourage a young doctor who does not wish to make statistics his profession, from learning to program on this type of software.

I can give you the example of one of my paediatrician residents who completed a Master 2 in biostatistics last year. She learned to use R, to perform data management and logistic regressions, and then submitted her paper. The reviewers ask him to carry out additional analyses. Today, she no longer understands her code, and needs the help of a statistician to move forward..

## Statistical software with a graphical interface

Some software is easier to use than R, SAS or Stata. The most famous being SPSS, which has drop-down menus to select the appropriate tests. It is indeed attractive to use this type of software, because it provides results, without having to learn to code. The counterpart, which is a major disadvantage, is that even if you do the wrong test, you will still get a result. And there is generally no way to check the validity conditions. In other words, they do not replace the rigour of statistical analysis. It is therefore mandatory to have very good statistical knowledge.

## TL;DR Does learning statistics give me a competitive advantage?

In highly competitive residencies, having published will very likely be beneficial to you. But it is important to distinguish between publications, statistics and statistical analyses. Good knowledge of statistics may be important depending on the carreer you want to build. Learning usual statistical software in order to perform statistical analyses by yourself will not benefit you. You could have used this time for something else. Your time is precious. Be efficient.

## Pvalue.io, the statistical software suitable for medical scientific research

A software is an exception in the world of statistical software : pvalue.io, designed for physicians and other health professionals who wish to publish. It does not require knowledge of statistics, but a research hypothesis or hypotheses. This software allows univariate, multivariate analyses to be carried out in a few clicks: linear and logistic regressions, and survival analysis. It is highly appreciated by doctors for its simplicity and effectiveness. pvalue.io provides the results in a Word file, and exports the figures in the formats requested by the medical journals (TIFF, EPS, etc). A test file gives you the opportunity to try it.

## No Comments