Data Skills for Scientists: Data and Navigating Uncertainty

Vocabulary flashcards covering foundational concepts from Forum 1 of Data Skills for Scientists, including data types, variables, lurking variables, paradoxes, accuracy versus precision, and principles of experimental and observational study design.

Data

Measurements or observations that have been directly made, which do not include any interpretation or context.

Information

Knowledge derived from data that incorporates interpretation, context, or calculations performed on raw data.

Variable

A characteristic or property of subjects that is measured or observed during data collection.

Categorical Variable

A qualitative variable often recorded as words, where the distance between values has no numerical meaning.

Numerical Variable

A quantitative, interval, or scale variable recorded as numbers, where the distance between values has meaningful interpretation.

Ordinal Variable

A categorical variable whose categories have a natural ordering (e.g., Olympic medals or letter grades).

Nominal Variable

A categorical variable whose categories do not have any natural ordering (e.g., car colours or ice cream flavours).

Discrete Variable

A numeric variable that takes values from a distinct, separate set of numeric values, typically representing a count.

Continuous Variable

A numeric variable that can take any value within a continuous interval of values (e.g., temperature, time, weight, or length).

Subject

The object, individual, case, or unit of observation on which measurements or observations are made.

Rectangular Data

A data structure where each variable is recorded for each subject, arranged such that rows represent subjects and columns represent variables.

Frequency

The count of how many subjects fall into each category of a categorical variable.

One-way Table

A table that summarises a single categorical variable by showing the frequency of subjects in each category.

Two-way Table

Also known as a contingency table; a table that summarises two categorical variables simultaneously by displaying counts for each combination of categories.

Aggregate Data

Information that summarises raw data by grouping and counting subjects across category combinations.

Accuracy

A measure closely associated with validity and interpreted as the opposite of bias; it describes how well data represents what it aims to measure on average.

Precision

A measure closely associated with reliability and interpreted as the opposite of variability; it describes how consistently data reproduces similar values under similar conditions.

Lurking Variable

A variable not included in a study that would dramatically alter the observed relationship between the studied variables if taken into account.

Simpson’s Paradox

A phenomenon in categorical data where an association consistent across all individual subgroups reverses or changes when all subgroups are combined.

Common Response

A situation where an observed association between two variables $X$ and $Y$ occurs because both change in response to a shared lurking variable $Z$.

Confounding

A condition where the effects of two variables (either explanatory or lurking) on a response variable cannot be distinguished from one another.

Descriptive Statistics

The branch of statistics concerned with summarising and visualising data to make its major features transparent.

Inferential Statistics

The process of drawing and presenting conclusions about a broader population based on sample data while quantifying uncertainty.

Observational Study

A study that observes subjects and measures variables of interest without intervening or assigning subjects to treatment groups.

Designed Experiment

A study that deliberately imposes specific treatments or conditions on subjects to measure responses and determine causality.

Treatment

The specific condition or set of conditions varied and administered to subjects by researchers in a designed experiment.

Response Variable

The outcome measured on subjects after treatments are applied, whose change in response to treatments is of primary interest.

Selection Bias

A bias that occurs when certain subjects are more likely to be included in a study than others, producing a non-representative sample.

Survivorship Bias

A form of selection bias where conclusions are drawn solely from subjects that passed an ordeal or selection process (such as returning combat aircraft).

Survey

A type of observational study where participants are asked to self-report information.

Response Rate

The percentage of individuals sent a survey who actually complete and return responses.

Control (Experimental Design)

The principle of experimental design that involves keeping unwanted sources of variation in the response variable constant or controlled.

Method of Comparison

An experimental framework where the outcomes of an active treatment group are compared directly against a control group.

Placebo

An inactive treatment that mimics the appearance, taste, or administration of an active treatment but contains no active ingredients.

Placebo Effect

A measurable change in the response variable observed in subjects receiving a placebo relative to an untreated control group.

Blinding

An experimental safeguard in which subjects are prevented from knowing which treatment they are receiving.

Double-Blind Study

An experiment where neither the participating subjects nor the experimenters administering the treatments know who receives which treatment.

Triple-Blind Study

A rigorous clinical trial design where subjects, treating experimenters, and data analysts evaluating the results are all kept unaware of treatment assignments.

Randomisation

The process of randomly allocating subjects to treatment and control groups to prevent selection bias and average out lurking variables.

Blocking

The experimental strategy of grouping subjects known to be similar in ways that affect the response, followed by random assignment of treatments within each block.

Matched Pairs Design

A randomized blocked experiment where each block consists of two closely matched experimental units or a single subject receiving both treatments in random order.

Replication

The practice of assigning a sufficient number of subjects to each treatment group to minimize the impact of chance variation.

Balanced Experiment

An experimental setup that allocates an equal number of subjects to each treatment group.