Data Analytics

Data analytics has become a key component of business practice. Companies rely on the essential measurements and findings to drive critical decisions about products, pricing, marketing, budgeting, resource allocation, and more. More complex analytics requires the expertise of a data analyst.

Data analysts examine, clean, and model data with the objective of highlighting business intelligence. Data analysis can be conducted with various techniques depending on the venue — business, science, education, government, or other areas.

Types of data analysis include:

  • Exploratory data analysis (EDA) for discovering new features in the data
  • Confirmatory data analysis (CDA) for confirming or denying existing hypotheses
  • Predictive analytics for forecasting or classification
  • Text analytics for classifying information

The Data Analysis Process

Data analysts typically practice data mining, an approach that focuses on modeling and knowledge discovery. The data analyzed may be quantitative (numerical) or qualitative (abstract).

The process of data analysis is conducted in phases:

  • Data Cleaning — Also referred to as scrubbing, this procedure involves identifying incorrect or misplaced bits of information in the data set and correcting or extracting as appropriate. This is critical, as eliminating outliers makes it possible to sort and organize data. However, a good data analyst will always note all changes to the data set and save previous versions, as a data scientist may possibly seek them out later.
  • Initial Data Analysis — In this phase, the analyst assesses the quality of data and the measurements used, notes extreme or outlying observations, compares differences in coding schemes, tests for common-method variance, and identifies the structure of the data set and any subgroups.
  • Main Analysis Phase — An exploratory or confirmatory approach can be adopted depending on what is pre-determined in the previous phase. Through cross-validation and sensitivity analysis, the reliability and stability of the results are determined. Statistical methods (general linear model, generalized linear model, structural equation modeling and/or item response theory) are put in practice.

At this point, the extracted, scrubbed, and analyzed data is compiled into a cohesive report. Different metrics are highlighted based on the stated goals of the business or institution. The report may be passed along to a data science team for further scrutiny.