# Workflow of Data Analytics

“Numbers have a compelling tale to share. They trust our ability to provide them with an authoritative and clear voice.” Stephen Few

Raw data aggregated is data that is not oriented. It requires a thoughtful understanding as well as the appropriate questions in order to create sense out of it. Many insights fail to analyse data completely and become difficult for the stakeholders’ comprehension. Therefore, it becomes necessary for a data analyst to define and understand data with the right set of initial questions and a standardized workflow for the different types of analysis he needs to perform.

The following words are from Jeff Leek’s fascinating book “The Elements of Data Analytic Style,” which broadly categorizes various analysis phases based on the type of question and the outcome expected to be achieved for the particular business need.

## Descriptive Data Analysis

The name suggests that this kind of analysis offers basic “descriptions” or summaries about the raw data set accumulated and the observations added to the same.

They can be both visual and quantitative, and the data can be depicted using statistics and simple graphs. This summary does not require any further analysis and is utilized as a summary to make sense of the information.

**Example:** Data on segregation of students enrolled in the same course at college:

The data could be split into various categories such as numbers, gender, residency age, race, and so on. The information summarizes or groups the data into a fixed set that describes all students and the specific information. It doesn’t suggest anything and only provides specifics. Thus, it is a type of descriptive analytics.

## Exploratory Data Analysis

Analysis of descriptive data output that is further studied for discoveries patterns, trends, correlations, or inter-relations among different areas of the data in order to develop an interpretation, an idea, or hypotheses. This is the foundation of Exploratory Data Analysis (EDA).

In essence, it’s expanding over the description data sets and trying to provide a comprehensive overview of the data. According to Dianne Cook, as well as Deborah F. Swayne rightly refer to in their book, “(EDA is) a ‘play-in-the-sand’ to allow us to find the unexpected and come to some understanding of our data.”

The main focus isn’t always the result of the problem statement; rather, to look at the various elements of data in the first place in order to more intimately.

**Example:** A typical EDA application studies the behaviour of traffic patterns in cities around the world. Although the data gathered may vary in terms of its nature, various surprising discoveries may be discovered like the frequency of accidents that occur at traffic signals, the amount of pollution that is produced on a daily basis because of exhaust emissions from vehicles, and even the rates of traffic congestion in a week. The outcome of the real issue isn’t always determined by these findings. The information gathered alongside other data may be helpful to determine the result.

## Inferential/Quantified Data Analysis

The distinction between inferential and exploratory analysis could be identified by determining if the analysis offers consistent information across various samples and the ones in the present.

**Example:** Calculating the mean of marks earned by students taking an exam against the difficulty index for 100 students can give valuable information on the students of 100.

This data can assist in understanding the quality of the connection between these two dimensions when studying student performance on exams. Although it’s impossible to know the reasons for these relationships, there is a way to determine the significance of a certain connection in determining inferential results.

## Predictive Data Analysis

The predictive analysis predicts the outcomes that could be expected from a small subset of data from the initial population set. This method of predicting new information is mostly built on quantifiable metrics from the existing data set.

Predictive analysis is not able to quantify the relationship between two dimensions as the inferential statistical method. Rather it uses probabilities that they share to predict possible outcomes in the future.

**Example:** Examining the influence and popularity of the nominees running for election to determine the outcome of that election.

In this case, we can determine the likelihood of the success of the candidate based on data about issues he discusses as well as his conservative and liberal views, information on his popularity in the state of his residence and so on. While we can estimate a potential outcome based on these data, however, we can’t predict the outcome accurately.

## Causal Data Analysis

Making modifications to one dimension or measurement to create a conclusive version of a different dimension is the foundation of causal analysis. It is designed to determine the extent and direction the measurement takes in contrast to the previous two. It is a predictive analysis and an inferential one.

**Example:** A randomized clinical trial to determine whether faecal transfer decreases the incidence of infections caused by Clostridium di-facile.

Patients in this research were randomly assigned to receive a faecal transfer along with standard care or regular treatment. Based on the results, the researchers found an unambiguous relationship between the outcomes of infections and transplants. Therefore, the study of the causality of patients produced an exact average outcome from raw data.

## Mechanistic Data Analysis

Although causal data provides an accurate average result, the aim isn’t just to comprehend that there’s an impact of the inferences derived from data but also to understand how the effect is affecting the outcome.

**An example:** Mechanistic analysis that examines the way in which wing design influences the flow of air around a wing, which results in less drag. In the absence of any engineering expertise, mechanical analysis of data is extremely difficult and is rarely done.

## Conclusion

As we can see, harnessing big-data analytics can bring huge benefits to companies, providing the context of data to tell an even more comprehensive story. By converting complex data sets into actionable intelligence, stakeholders can make better business decisions. If we know how to make big data accessible to our clients, the value of our service is now ten times greater.