Top Page | Upper Page | Contents | About This Site | JAPANESE

Computational Statistics

Computational statistics is a discipline that combines computer simulation and statistics.

There are advanced methods, but even introductory methods of computational statistics, such as "creating scattered data and verifying formulas", It can help you unravel real-world problems. I myself have a number of experiences.

"Even the introductory method of computational statistics alone can help solve real problems" is not widely recognized, but it is a promising field of data science.

Uses of Computational Statistics

The following are introductory uses of computational statistics.

See the relationship between theoretical formulas and data

Statistical formulas often contain error terms. This allows you to handle variations.

If you use How to create variable data, you can create data like the example of a a href="ede1-6-3.html">Dispersion Model, and you can ask "Is it close to reality?" or "Can it be used as a real model?" You can do a survey that.

Try again and again

The actual variability in data is something that "accidentally" appears at that time. I only know about one dose.

"Did it happen to be terrible at that time?" or "Does it happen often?" You can investigate that. The Random Walk Model page is an example.

Also, it is a well-known story that the mean value of the variance becomes an unbiased variance, but "What is the median?" and "How much variation does it have when the number of samples is small?" Such as, It is difficult to handle with a theoretical approach. The Unbiased Variance page examines the characteristics of unbiased variance from the perspective of computational statistics.

Research and deepen understanding of statistics

Real-world data is essential for real-world research because it contains unique information and context.

one side When researching theories in statistics or studying theories, the information held by real data may become noise, and the important points may not be visible.

If you use "data that is uniformly distributed" or "data that is normally distributed and the mean and standard deviation are known", You will be able to conduct research and study without noise. In addition, when dealing with real data, it is easier to distinguish between the parts that can be estimated from theory and the parts that are unique to real data.

Computational Physics

Computational physics is a field similar to computational statistics.

Traditional physics is divided into theoretical physics and experimental physics. In physics, advanced mathematics is sometimes used, and theoretical physics is the study of what the mathematical part is. Theoretical physics researchers will be working with only pen and paper. On the other hand, experimental physics elucidates physical phenomena through experiments.

At my alma mater, there were half laboratories for theoretical physics and half experimental physics.

Computational physics is a field that rapidly developed as the "third type of physics" in the 1990s, when I was a university student. It is the same as theoretical physics in that it does not deal with actual "objects", but it also has an aspect of experimental physics because it conducts experiments (simulations) in computers.

When I was in the master's program, I was fascinated by this area, and with an awareness of environmental problems, I proceeded to research on Lead Free and First Principle Calculation.

NEXT Multi-Variable Analysis