Types of Statistical Analysis

Univariate analysis is one of the primary forms of data analysis techniques. If we break down the term, “uni” means one and “variate” indicates variable. This means that it deals with only one variable and doesn’t consider the cause-and-effect relationship. As a result, we only explore each variable separately and don’t consider what the effect of variable x is on variable y. Univariate analysis helps us find patterns within data. We can perform it for both categorical and numerical variables.

We can use a variety of central tendency measures, including mean, median, mode, dispersion, and others, to find patterns within data. In addition, we can visualize univariate data using pie charts, histograms, bar charts, and so on.

Let’s say we record the score for six players who played a cricket match. The scores are given in the table below:

Here, we are observing only one variable, Score. We can derive many patterns just by following the Score variable. For example, we can see that the highest Score is 100. Similarly, the minimum Score is 30. The average Score in this match is 55.83.

Let’s discuss another scenario—we have data for an ice cream parlor that displays the frequency a specific flavor is ordered, as shown in the table below. Here, we are only observing the Orders variable for each ice-cream flavor.

By observing the number of Orders for each flavor, we can quickly determine which flavor is popular among customers. We’ve represented this using the pie chart below. We can see that chocolate is the most popular flavor among customers.

“Bi” means two, and “variate” indicates variable. In a bivariate analysis, as the name implies, we compare two variables and study the cause-and-effect relationship. We can observe how the change in one variable can affect the other variable.

Let’s say we have some data that represents the number of Hours students studied for a final exam and the final Scores earned, shown in the table below:

Here, we are comparing two variables. Therefore, it is a bivariate analysis. If we closely observe the patterns in the data, we can conclude that students who spent more hours studying for the exam earned higher scores. As a result, the number of Hours and Scores are directly proportional—if we increase the number of Hours, the Scores also increase. If we represent the Scores earned and the number of Hours studied in the x-y plane, we can see the pattern shown below:

As illustrated in the graph, we’ve plotted Hours on the x-axis and Scores on the y-axis, and we can see an increasing pattern. Here, we’ve compared the numeric variable Hours with the numeric variable Scores and the resulting graph is a scatter plot. We can perform bivariate analysis on both numerical and categorical variables. We can also use regression plots and correlation coefficients for bivariate analysis.

In multivariate analysis, we compare three or more variables and study each variable’s effect on the other. It’s similar to bivariate analysis, but in this case, more variables are involved.

Consider a scenario where a team of researchers wants to assess the cancer risk in a given community. Although it is generally accepted that the risk of having cancer rises with age, we cannot make any inferences based just on this assumption (univariate analysis).

Similarly, another study suggests that smoking may also increase cancer risk. However, relying solely on these two variables (bivariate analysis) may not be sufficient to provide an answer to our question. It is a complex problem, and we need to consider various factors, such as age, family history of cancer, smoking habits, obesity, a poor diet, a sedentary lifestyle, and so on. So, in situations like that, multivariate analysis is the right choice.

Let’s say a marketing team is studying how consumers decide which things to buy. Naturally, the pricing element is the first factor to consider because customers are attracted to reduced prices. However, if we only consider the price factor, it might provide insight into what customers enjoy. Still, one approach won’t work for everyone because consumer tastes are varied and complex. Numerous factors need to be taken into account, including the product’s brand, quality, accessibility, as well as social and psychological aspects, to name a few.

So in a situation like that, multivariate analysis is the appropriate choice. Multivariate analysis is needed when the problem at hand is complex and necessitates a much more in-depth analysis of the data. In other words, we use multivariate analysis when relying on just one or two variables is insufficient and we need to consider numerous aspects in order to solve the problem at hand.