Data Collection in Statistics

Introduction to Statistics

Statistics is the science of collecting, analyzing, and describing data, often with the goal of making conclusions about a larger group known as the population. The process of data collection is foundational, as it determines the quality and reliability of any subsequent analysis or inference.

• Population: The entire group of interest in a study (e.g., people, countries, animals).

• Sample: A subset of the population from which data are actually collected. The size of the sample is denoted by n.

• Case/Unit: An individual member of the sample.

• Census: Data collected from every member of the population.

Example: If a study observes n = 6000 people using a public washroom, the sample consists of those 6000 individuals.

Variables and Data Values

A variable is a characteristic recorded for each case in a study. Variables can be classified as either categorical or quantitative:

• Categorical Variable: Classifies cases into categories (e.g., Yes/No, race, marital status).

• Quantitative Variable: Takes numerical values for which arithmetic operations like addition and averaging make sense (e.g., age, income, body mass).

The value of a variable is the specific measurement or category recorded for a case. The set of all values for a variable across cases is called the values of the variable.

Example: In a dataset of mice, the variable BodyMass might have a value of 15g for mouse #7.

Data Tables

A data table is a two-dimensional display where rows represent cases and columns represent variables. The cell entries are the values of the variables for each case.

Example Table Structure:

Classifying Variables

• If the values are words or categories, the variable is categorical.

• If the values are numbers representing measurements or counts, the variable is quantitative.

• Numbers can sometimes represent categories (e.g., 1 = Yes, 0 = No).

• Quantitative variables can be grouped into categories to create a categorical variable.

Example: The variable Smoke (Yes/No) is categorical, even if we count the number of "Yes" responses.

Example: Developed (1 = underdeveloped, 2 = developing, 3 = developed) is categorical, even though it is based on a quantitative measure (energy usage per capita).

Units and Categories

• For categorical variables, always list the possible categories.

• For quantitative variables, specify the units (e.g., years, dollars, grams).

How Data Are Collected

Sampling Methods

• Convenience Sample: Data collected from cases that are easily available. Not scientifically reliable for generalizing to a population.

• Anecdotal Evidence: Data based on personal experiences; not considered scientific evidence.

• Random Sample: Each member of the population has a known chance of being selected, minimizing sampling bias.

Sampling Bias occurs when the method of selecting the sample causes it to differ from the population in a relevant way. Types include undercoverage, self-selection (voluntary response), and nonresponse.

Response Bias occurs when the responses given by cases differ from the truth. Examples include:

• Wording Bias: The phrasing of questions influences responses.

• Social Desirability Bias: Respondents answer in a way that is viewed favorably by others.

Key Point: Random sampling is essential for making valid inferences about a population.

Sample Size

• A small random sample is more reliable than a large non-random sample.

• Large samples are beneficial but not a substitute for randomness.

Relationships Between Variables

Association and Causation

Two variables are associated if knowing the value of one helps predict or explain the value of the other. The variable used for prediction is the explanatory variable, and the variable being predicted is the response variable.

• Association does not imply causation.

• Possible explanations for association:

  • Explanatory variable causes changes in the response variable.

  • A confounding variable influences both variables.

  • Coincidence (chance).

Types of Studies

• Observational Study: Researchers observe variables in their natural state without intervention.

• Experiment: Researchers manipulate the explanatory variable (treatment) and observe the effect on the response variable.

It is difficult to establish causation in observational studies due to potential confounding variables.

Randomized Experiments

• Randomized Experiment: The value of the explanatory variable is assigned randomly to each case before measuring the response.

• If a randomized experiment shows an association, a causal relationship can be established.

• Randomized Comparative Experiment: Cases are randomly assigned to different treatment groups for comparison.

• Matched-Pairs Experiment: Each case receives both treatments in random order, or cases are paired and each receives a different treatment. This design reduces variability.

Controls and Blinding

• Control Group: Receives a treatment that does not directly affect the response variable (e.g., placebo).

• Placebo Effect: Improvement due to the belief in the treatment rather than the treatment itself.

• Blinding: Participants (and sometimes researchers) do not know which treatment is being administered to prevent bias.

• Double-Blinding: Both participants and researchers are unaware of treatment assignments.

Replication of results across multiple experiments or cases increases scientific confidence in findings.

Summary Table: Types of Variables

Key Formulas and Notation

• Sample Size:

• Proportion (relative frequency):

Conclusion

Understanding how data are collected, the types of variables, and the design of studies is essential for making valid statistical inferences. Random sampling and well-designed experiments are crucial for drawing reliable conclusions about populations and for establishing cause-and-effect relationships.