JNTU B.Tech III Year DATA ANALYTICS Previous Question Paper (February - 2022 - R18)

Market researchers utilize experimental designs to establish cause-and-effect relationships by manipulating independent variables and observing their impact on dependent variables. The four primary experimental designs are:

1. Pre-Experimental Designs: These designs lack a control group or randomization, making them weak for establishing causality. Examples include the One-Shot Case Study, where a single group is exposed to a treatment and measured once, and the One-Group Pretest-Posttest design, which measures the group before and after the treatment.
2. True Experimental Designs: These are the gold standard because they involve random assignment of test units to experimental and control groups. The Pretest-Posttest Control Group design measures both groups before and after the treatment, while the Posttest-Only Control Group design measures only after the treatment, effectively controlling for most extraneous variables.
3. Quasi-Experimental Designs: These are used when full randomization is not feasible, such as in field settings. The Time Series design involves multiple measurements of the dependent variable both before and after the intervention, allowing researchers to observe trends over time. The Nonequivalent Control Group design is similar to the true experimental pretest-posttest design but lacks random assignment.
4. Statistical Designs: These allow for the simultaneous manipulation of multiple independent variables and the control of extraneous variables. The most common is the Factorial Design, which measures the effects of two or more factors at various levels. This design helps identify interaction effects between variables, providing a more comprehensive understanding of complex market dynamics.

Data quality refers to the condition of a set of values of qualitative or quantitative variables. High-quality data is accurate, complete, consistent, timely, valid, and unique. It is a critical prerequisite for meaningful data analysis, as poor-quality data leads to unreliable models and incorrect business decisions.

Data preprocessing is the essential phase in data analytics that transforms raw, noisy, and incomplete data into a clean and structured format suitable for mining or modeling. The primary steps involved in data preprocessing are:

1. Data Cleaning: This involves handling missing values (via imputation or deletion), smoothing noisy data (using binning or regression), and identifying or removing outliers to ensure consistency.
2. Data Integration: This step merges data from multiple heterogeneous sources (such as databases or files) into a coherent data store, while resolving conflicts in data representation and schema matching.
3. Data Transformation: Data is converted into appropriate forms for mining. Techniques include normalization (scaling data to a specific range like [0, 1]), discretization (converting continuous data into intervals), and attribute construction (creating new features from existing ones).
4. Data Reduction: This reduces the volume of data while maintaining the integrity of the original information. Techniques include dimensionality reduction (e.g., Principal Component Analysis), numerosity reduction (e.g., sampling or clustering), and data compression.

Effective preprocessing significantly improves the efficiency of analytical algorithms and the accuracy of the resulting predictive models.