Biomedical Research: Variable Types & Data Analysis

• Session learning objectives include outlining different variable types in biomedical research.
• Session learning objectives include determining the applicability of different summary statistics for descriptive data analysis.

Types of Variables (Categorization in the UK)

• Variables are categorized as either numeric or categorical.
• Numeric variables can be continuous or discrete.
• Continuous variables can take any possible value within a given range, like temperature in Celsius.
• Discrete variables can take specific values within a given range with gaps in the range of variables.
  • A typical discrete variable is a count of integers, such as the number of teeth.
• Categorical variables can be normal, ordered, or binary.
• A normal categorical variable is divided into different categories not ordered in any way, like the brand of a car.
• An ordered categorical variable is divided into different categories in an ordered manner, like educational level (elementary school, high school, university).
• Binary categorical variable has only two categories, such as affected by the disease (yes vs no).

Types of Variables (Categorization in the US)

• In the US variables are also categorized as Numeric and Categorical; however in the following subcategories
• Numeric variables are divided into interval scale.
• Interval scale: values are graded in equal increments along the whole range of values, e.g. temperature in Celsius.
• A numeric variable can also be ratio scale.
• Ratio scale: Similar to interval scale, but a true zero exists, indicating a true absence of quantity, e.g., temperature in Kelvin.
• Categorical variables are divided into nominal scale.
• Nominal scale: divided into different categories that are not ordered, e.g., country of origin.
• A categorical variable can be an ordinal scale, such as income categorized as "low income", "middle income", or "high income".

Examples of Variables

• Blood triglyceride levels
• Dietary pattern can be categorized as very healthy, moderately healthy, moderately unhealthy, or very unhealthy.
• Cyprus district can be Nicosia, Limassol, Larnaca, Paphos, Famagusta, or Kerynia.
• Number of asthma attacks per month.
• The type of variable depends on its definition and how it's used in a study.
• A blood pressure variable can be defined as hypertensive (yes vs no), which is binary categorical.
• If blood pressure is defined in mm Hg, it is numeric continuous.

Summarizing Categorical Variables: Proportions

• Actual (absolute) numbers are rarely informative and may be misleading.
• Most cars involved in accidents in New Zealand are white.
• There are 900,000 diabetics in Greece and 1,800,000 diabetics in Germany.

Basic Summary Statistics for Categorical Variables

• Categorical variables are presented by counting data values in each category of interest.
• The distribution of categorical variables can be organized into a frequency table showing the numbers and proportion for each category.
• For example, with pulmonary tumors, 32 cases were benign (80%), and 8 cases were malignant (20%) out of 40 total cases.

Presenting Categorical Data with Graphs

• Categorical data can be represented using bar charts and pie charts.
• A bar chart displays the distribution of a categorical variable, showing the counts for each category, shown side by side for comparison.
• Bar charts have same widths and keep equal spaces between bars.
• Pie charts display the total population/sample as a cicle.
• They slice the circle into pieces, with slices proportional to the fraction of the whole in each category.

Homework

• Create a dataset containing the following variables: Smoking status and Height.
• An exercise using Excel is to summarize the variable ‘smoking status’ and create a bar chart.
• An exercise using Excel is to summarize the variable ‘smoking status’ and create a pie chart.
• For calculations, calculate the frequency of each observation using the Excel function 'COUNTIF'.
• Convert the result to a proportion by dividing by the total number of individuals.