Data Foundations and Classifications

In the study of statistics, understanding the nature and structure of data is fundamental. This chapter explores the distinction between populations and samples, defines what constitutes data, and provides a comprehensive classification system for the various types of information statisticians encounter.

I. Populations and Samples

Before data can be analyzed, researchers must define the scope of their study. A population is defined as the total collection of all elements under study about which a researcher intends to draw conclusions. For instance, if a study focuses on voting behavior, the 60 million eligible voters would constitute the population. Conversely, a sample is a representative portion of that population. A common example of sampling is blood testing, where a small amount of blood represents the entire biological system.

II. Defining and Classifying Data

Data refers to information collected in a systematic manner to achieve specific research objectives. Linguistically, "data" is the plural form of the word "datum". Data can be classified based on five primary criteria: its source, the type of research, the method of measurement, the time frame, and the scale of measurement.

1. Classification by Source

• Primary Data: This consists of original information generated through large-scale surveys, often conducted by government bodies, individuals, or research institutions. A population census is a classic example of primary data.
• Secondary Data: This refers to information that was not originally collected by the current researcher but obtained from existing published or unpublished sources. Examples include data from the Bangladesh Bureau of Statistics (B.B.S.) or the Economic Survey of Bangladesh.

2. Classification by Research Type

• Survey Data: This is collected through direct methods such as interviews, questionnaires, or direct observation.
• Experimental Data: This is information specifically obtained through controlled experiments.

3. Classification by Measurement

• Qualitative Data: This information is classified based on attributes or qualities that cannot be measured numerically, such as gender or hair color.
• Quantitative Data: This refers to characteristics that are numerically measurable, such as height, weight, income, sales, or production figures.

III. Data Structures Over Time

The timing of data collection significantly impacts how it is analyzed. There are four major temporal classifications:

• Time Series Data: This involves observations of one or more variables collected over a sequence of time. Common examples include stock prices, GDP, and annual crime rates.
• Cross-Sectional Data: This is a dataset collected by sampling a population (such as individuals, firms, or cities) at a single, specific point in time.
• Pooled Cross Sections: This configuration combines independent cross-sections collected at different points in time into a single dataset.
• Panel (Longitudinal) Data: This is the most complex structure, consisting of a time series for every cross-sectional member in the dataset, such as tracking crime statistics for the same set of cities over multiple years.

IV. Levels of Measurement (Scales)

Variables generally fall into four scales of measurement, divided into numerical and categorical categories.

Numerical Data Scales

• Ratio Scale: This is the highest level of measurement where ratios, distances, and natural ordering are all meaningful. Most economic variables, such as GDP growth percentages, fall into this category.
• Interval Scale: On this scale, the distance between categories is known and constant, and natural ordering exists. However, unlike the ratio scale, the ratio between two values is not meaningful; for example, the distance between years (2005 to 2006) is meaningful, but the ratio of the years themselves is not.

Categorical Data Scales

• Ordinal Scale: This is used when there is a natural ordered relationship among categories (e.g., "higher," "greater," or "less difficult"). While the order exists, the precise distance between categories cannot be quantified, such as in letter grading systems (A, B, C) or income classes (Upper, Middle, Lower).
• Nominal Scale: This is a figurative labeling scheme where numbers serve only as tags for classification. It possesses none of the features of a ratio scale, with ID numbers being a typical example.

V. Methods of Data Collection

Statisticians utilize both formal and informal methods to gather information.

• Formal Methods: These include structured approaches like interviews (personal, telephonic, or electronic), questionnaires, direct observation, and designed experiments.
• Informal Methods: These often involve more qualitative approaches, such as focus group discussions, depth interviews, and projective techniques.