Using Statistical Regression Methods in Education Research

SRME

Glossary

I

Independent errors

This is an assumption of regression techniques that states that residuals (error terms) for cases should not be correlated. We do not cover this in depth on this website, partly because of its complexity, but mainly it is usually only something to be concerned with if you are collecting data which has a clear hierarchical structure. See SLR module 2.6 for further detail.

Inferential statistics

Statistical analyses which allow us to draw inferences about a given population based on data taken from an appropriate sample. Theoretically, if certain stringent conditions are met, information about the population can be extrapolated by gathering data from a smaller but representative group of that population.

Interaction effect

An interaction effect occurs when the impact of one variable on an outcome is not the same at all levels of a second explanatory variable. We discuss examples in SLR module 3.11-3.13. For example Bangladeshi students may have lower achievement than White British students among those from advantaged (high SES) backgrounds, but from higher achievement than White British students among those from disadvantaged (low SES) backgrounds. So the 'effect' of ethnicity varies depending on the level of SES i.e. there is an interaction between ethnicity and SES in their effects on academic attainment. Evaluating interactions between explanatory variables is often crucial to building accurate regression models.

Intercept

The intercept is the predicted value of the outcome variable in a regression model when all of the explanatory variables have a value of zero. If this was displayed on a scatterplot for a simple linear regression it would be the point at which the regression line crosses, or intercepts, the zero point on the Y-axis (the axis for the outcome variable). See SLR module 2.5 for example.

Interval variable

Interval variables constitute a type of data that falls under the umbrella of Scale data in SPSS. Intervals along the scale of measurement for the variable must be equal. For example, imagine you are measuring attitude to homework on a scale of 1 to 10 such that 1 = 'I hate home work' and 10 = 'I love home work' (this response sounds sarcastic to us but maybe some students do!). For the variable to be Interval the difference in attitude between a score of 1 and 2 needs to be the same as the difference between 4 and 5 and the same as the difference between 7 and 8, etc. In other words all of the intervals represent the same difference in attitude.

Page contact: Feedback to ReStore team Last revised: Thu 28 Jul 2011
Back to top of page