Analytics for Decision Making Glossary

Alternative hypothesis

The hypothesis for which one collects evidence to prove.

Bar chart

A visual display for categorical data.

Binomial distribution

A probability distribution based on exact known probabilities of an event.

Categorical data

Data that is non-numeric, such as color.

Confidence level

The desired percent level for which one wishes to be accurate.

Confidence interval

An interval based on sample data to estimate an unknown population parameter.

Continuous data

Data that is measured, such as height or temperature.

Convenience sample

A sample collected from 'family and friends' that may not be representative of the population.

Coverage error

When a population frame from which a sample is taken does not accurately represent the population.

Cumulative frequency distribution

A tally of values from the lowest value in a data set to the upper limit of a class interval.

Discrete data

Data that is counted, such as cell phones or people.

Empirical rule

A guideline to estimate the percentage of data that falls between 1, 2, & 3 standard deviations in normally distributed data.

Frequency distribution

A tally of values for categorical data.

Histogram

A visual display for continuous data.

Hypothesis test

A statistical test in which sample data is collected to determine if there has been a change from a previous assumed numeric value.

Mean

The mathematical center or average of a data set.

Measurement error

An error that occurs when incorrectly recording data from a sample.

Non-response error

An error that may occur due to the fact that not all surveyed individuals respond.

Normal distribution

A continuous probability distribution where most data is symmetrically distributed around the mean.

Null hypothesis

The assumed case of a hypothesis for which one collects evidence to disprove.

Numerical data

Data that can be measured or counted.

p-value

The probability of obtaining a sample statistic as or more extreme than the one observed, assuming the null hypothesis is correct.

Point estimate

The center value of a sample likely used in a confidence interval or hypothesis test.

Population mean

The mathematical center or average of a data set for a large population.

Population proportion

The mathematical center or average of a data set for a large population.

Primary data

Data that is directly collected by the person analyzing it.

Qualitative data

Data that is non-numeric, such as color (see categorical data).

Quantitative data

Data that can be measured or counted (see numeric data).

Random sample

A sample taken based on pure random selection.

Sample mean

The mathematical center or average of a sample.

Sample proportion

The mathematical center or average of a sample.

Sample size

The number of observations in a sample.

Sampling error

An error whereby the sample is not representative of the population due to random chance.

Secondary data

Data that is collected by someone other than the person analyzing it.

Stratified random sample

A sample where subgroups are created, and a random sample is taken from each subgroup.

Standard error

The measure of variation around a sample statistic, such as a sample mean or sample proportion.

Standard deviation

A measure of variation for a data set, thought of as the average deviation from the mean.

Statistics

Numeric measures that describe a data set for a sample.

Summary table

A table showing the number of occurrences of an item, usually for categorical data.

t-value

The number of standard deviations or standard errors that a data point lies from the mean, used when the population standard deviation is unknown.

Type I error

The error resulting from rejecting a null hypothesis when it is true.

Type II error

The error resulting from not rejecting a null hypothesis when it is false.

Uniform distribution

A distribution where each data value is equally likely to occur.

Variable

A characteristic about an item in a data set.

Z value

The number of standard deviations or standard errors that a data point lies from the mean.

Anchoring

Presenting subtly leading data before asking respondents to make a quantitative judgment.

Conjunction fallacy

The tendency to believe that more complicated scenarios are more realistic than simpler ones.

Priming

Influencing a quantitative judgment by asking a respondent to think about a particular topic.

Dependent Variable

A variable, usually denoted by Y, which is being predicted in supervised learning.

Independent variable

A variable, usually denoted by X, used as an input into a predictive model.

Mean Absolute Percentage Error (MAPE)

A metric that shows the average error made relative to the true values, in percentage terms.

Residual

The difference between a datum and the value predicted for it by a model.

Training Set

The portion of the data used to fit a model.

Supervised Learning

The process of providing an algorithm with records where an output variable is known.

Validation Set

A sample of data not used in fitting a model, used to assess model performance.

Analytics

The extensive use of data, statistical, and quantitative analysis, models, and management to drive decisions.

Cohort

A group of customers who share something in common.

Cohort Analysis

Temporal study of user data to gain insights on behavior over time.

Customer Segmentation

Dividing users based on behavior, demographic, past purchase activity, etc.

Customer Retention

Actions taken to encourage customers to revisit and reduce churn rates.

RFM Segmentation

A method for scoring/segmenting customers based on recency, frequency, and monetary value.

Cluster Profiling

Describing clusters by the average and range of variables in each cluster.

Segmentation Modeling

Applying a unique model to each data cluster or subset.

A/B testing

A method for testing different webpage versions to see which produces the best outcome.

Conversion Rate

Ratio of total users who take a desired action over all visitors.

Type I Error (false positive) in A/B Testing

Concluding that variant A is a winner when it does not perform better than variant B.

Type II Error (false negative) in A/B Testing

Concluding that variant A is no different from variant B when it actually is.