Top Page | Q&A | About This Site | JAPANESE



Why don't you just swap the rows and columns?

Background of the question

I explain Analyzing variable companioning and Analysis of Similarity of Samples when classifying data analysis methods .

So, "Although it is divided into two types, if you swap the rows and columns (transpose), you can analyze the grouping of variables by the method of analyzing the grouping of samples, and vice versa? It is the idea of ??the question.


The answer to the general case is no. The answer for special cases is "yes".

First, I explain the general case.

The method of analyzing variable grouping can handle a collection of variables with different physical meanings and units such as "height, weight, temperature, and humidity". If you swap the rows and columns of such a matrix and analyze the grouping of the samples, you will get "There is nothing that is completely different from each other and has a grouping relationship."

On the other hand, when you want to analyze the grouping of samples for data in which variables with various meanings are mixed like this, you can replace the rows and columns and use the method of analyzing the grouping of variables. There is nothing different in one group, "and I can't find out what I want to find out.

Next, I explain a special case.

A special case is a case where all variables have the same physical meaning and unit, unlike the general case above. There are data in the world where only the number of occurrences (frequency) is collected and data where only the amount of money is collected, but such data is used.

With such data, it is possible to swap rows and columns, as it does not happen in the general case above.

By the way, with such data, you can also use Analysis of Many vs Many and Analysis of similarity between items in rows and columns on this site .