Term	Description
Response variable	Gene expression, e.g. log-CPM values.
Explanatory variable	Variable that influences gene expression.
Statistical model	Used to describe the relationship between response and explanatory variables.
Model parameters	Of statistical models, unknown but estimable values that describe the direction and magnitude with which explanatory variables influence gene expression.
Design matrix	Used to define the form of a statistical model and to store observed values of the explanatory variable(s). Used in the computation process to estimate model parameters.
Contrast matrix	Used in conjunction with a design matrix to calculate specific values of interest between estimated parameters.
Covariate	Explanatory variable that is numerical in nature, e.g. age.
Y-intercept	Point at which a model prediction crosses the y-axis.
Slope	Rate of change for a model e.g. the change in gene expression per unit increase of a covariate.
Regression model	Our reference to statistical models for covariates.
Factor	Explanatory variable that is categorical in nature, e.g. genotype.
Levels	Unique values within a factor, e.g. wildtype or mutant.
Means model	Our reference to statistical models for factors where parameters are calculated as the mean of each factor level.
Contrasts	Linear combinations of estimated parameters. A contrast matrix is made up of individual contrasts.
Mean-reference model	Our reference to statistical models for factors where parameters are calculated as the mean reference level, and relative means for subsequent levels.
Fitted model	The statistical model written with estimates of the model parameters. In our figures, we draw the fitted model (expected gene expression) along with the data points (observed gene expression) to give an idea of how well the fitted model represents the relationship between response and explanatory variables.
Additive effect	When the combined effect of two factors equals the sum of the two individual effects, e.g. if the estimated effect of Group A is κ and the estimated effect of sequencing on lane I is τ, then a sample in Group A that is sequenced on lane I has an expected expression of κ + τ if the two factors have an additive effect.
Interaction effect	When the combined effect of two factors does not equal the sum of the two individual effects, e.g. for the example above, a sample in Group A that is sequenced on lane I has an expected expression of κ + τ + δ if the two factors have an interaction effect, where δ can be a positive or negative number.
Nested factors	A factor is considered to be nested within a second factor, e.g. group is nested within batch, if different sets of its levels can be found in each level of the second factor, e.g. group A and group B are processed in batch B1 and group C and group D are processed in batch B2.
Mixed effects models	A statistical model that contains both fixed and random effects, where random effects are usually not of interest to the study at hand.