Investigative Process
In the very first step of a statistical study, you must create one of these to define exactly what you want to learn.
What is an investigative question?
Asking only the friends sitting at your lunch table because they are easy to reach is an example of this highly biased sampling method.
What is a convenience sample (or convenience sampling)?
Unlike an observational study where researchers merely measure existing conditions, a study is only considered an experiment if researchers deliberately impose one of these on the subjects.
What is a treatment?
When the difference observed between two groups is far too large to plausibly explain by random variation alone, researchers declare the result to be this.
What is statistically significant?
While height and weight are quantitative variables measured in numbers, eye color and state of birth belong to this data type based on labels or attributes.
What is categorical data (or qualitative data)?
To quickly collect data from a large group of people about their opinions or habits, researchers often hand out this written tool.
What is a survey?
Online polls and call-in radio shows often suffer from this bias because only people with strong opinions choose to participate.
What is voluntary response bias?
This key principle of experimental design uses chance to assign subjects to treatment groups, ensuring that baseline differences and unknown traits are spread out evenly.
What is random assignment (or randomization)?
By standard scientific convention, a result is typically declared statistically significant if the likelihood of it happening purely by chance falls below this percentage threshold.
What is 5% (or 0.05)?
This circular graph displays categorical data where each section represents a proportion of a whole, requiring all slices to sum to 100%.
What is a pie chart (or circle graph)?
When arranging a set of numbers in order from smallest to largest, this term refers to the number sitting in the exact middle.
What is the median?
Conducting a survey using only landline phone numbers causes this type of bias by leaving out young adults and mobile-only households from the sampling frame.
What is undercoverage?
To prevent outside factors—like passing time or seasonal changes—from skewing results, researchers compare a treatment group against this baseline group that receives an inactive treatment or no treatment at all.
What is a control group?
In a study comparing two prep courses, Group A scored 8% higher than Group B. Researchers did this and found there was a 0.03 chance the 8% difference occurred strictly by random luck.
What is a simulation?
Ideal for small quantitative datasets, this display stacks a mark for each observation above a number line, preserving every exact individual data value.
What is a dot plot?
Researchers often plot their data visually using simple diagrams to show counts for different categories in this step of the investigative process.
What is analyze data?
This bias occurs when individuals chosen for a study refuse to participate or cannot be contacted, which skews results if their opinions differ from those who responded.
What is nonresponse bias?
Testing a new medication on just two people isn't enough; experimenters rely on this principle by using many subjects in each group to reduce the impact of chance variation.
What is replication?
To accurately generalize survey results from a small group to an entire population without selection bias, researchers must select participants using this method.
What is random sampling (or a random sample)?
This graph displays continuous quantitative data using touching bars built over equal-width numerical intervals called bins.
What is a histogram?
When drawing conclusions from your results, statisticians always remind you that just because two things are correlated, it does not prove this.
What is causation (or cause and effect)?
Misleading question wording, interviewer influence, or a participant lying to give a socially desirable answer all lead to this type of bias.
What is response bias?
Observational studies can reveal strong associations, but they cannot prove a cause-and-effect relationship because they fail to prevent these unmeasured variables from interfering with the response.
What are confounding variables (or lurking variables)?
In an experiment, this practice evenly spreads out baseline differences and confounding variables across treatment groups, allowing researchers to draw a direct cause-and-effect conclusion.
What is random assignment?
Recorded entirely in numbers, zip codes, athlete jersey numbers, and phone numbers are classified as this variable type.
What is a categorical variable (or categorical data)?