CHAPTER 15 Introducing Correlation and Regression 209

alphabet (a, b, c, d). There’s no consistent rule regarding uppercase versus lower-

case letters.

Sometimes a collection of predictor variables is designated by a subscripted vari-

able (X

X

1

2

,

and so on) and the corresponding coefficients by another subscripted

variable (b

b

1

2

,

, and so on).

In mathematical texts, you may see a regression model with three predictors

written in one of several ways, such as»

» ^Z

a

bX

cY

dV (different letters for each variable and parameter)»

» ^Y

b

b X

b X

0

1

1

2

2 ^{(using a general subscript-variable notation)}

In practical work, using the actual names of the variables from your data and

using meaningful terms for parameters is easiest to understand and least error-

prone. For example, consider the equation for the first-order elimination of an

injected drug from the blood, Conc

Conc

e ^k

Time

e

0

. This form, with its short but

meaningful names for the two variables, Conc (blood concentration) and Time

(time after injection), and the two parameters, Conc0 (concentration at Time 0)

and ke (elimination rate constant), would probably be more meaningful to a reader

than Y

a

e ^b

X

-

.

Classifying different kinds of regression

You can classify regression on the basis of»

» ^{How many predictors or independent variables appear in the model»}

» ^{The type of data of the outcome variable»}

» ^{What mathematical form to which the data appear to conform}

There are different terms for different types of regression. In this book, we refer

to regression models with one predictor in the model as simple regression, or uni-

variate regression. We refer to regression models with multiple predictors as mul-

tivariate regression.

In the next section, we explain how the type of outcome variable determines

which regression to select, and after that, we explain how the mathematical form

of the data influences the type of regression you choose.