2 Algebra Review
2.1 Overview
Algebra, arthmetic, and probability are underlying frameworks for statistics. The following elements of algebra are particularly important:
Understanding symbols as variables, and what they can stand for
Factoring out common terms: \(axw + bx = x(aw + b)\)
Factoring out negation of a series of added terms: \(-a - b = - (a + b)\)
Simplification of fractions
Addition, subtraction, multiplication, and division of fractions
Exponentiation with both fractional and whole number exponents
Re-writing exponentials of sums: \(b^{u + v} = b^{u}\times b^{v}\)
Logarithms
- log to the base \(b\) of \(x\) = \(\log_{b}x\) is the number \(y\) such that \(b^{y} = x\)
- \(\log_{b}b = 1\)
- \(\log_{b}b^{x} = x \log_{b}b = x\)
- \(\log_{b}a^{x} = x \log_{b}a\)
- \(\log_{b}a^{-x} = -x \log_{b}a\)
- \(\log_{b}(xy) = \log_{b}x + \log_{b}y\)
- \(\log_{b}\frac{x}{y} = \log_{b}x - \log_{b}y\)
- When \(b = e = 2.71828\ldots\), the base of the natural log, \(\log_{e}(x)\) is often written as \(\ln{x}\) or just \(\log(x)\)
- \(\log e = \ln e = 1\)
Anti-logarithms: anti-log to the base \(b\) of \(x\) is \(b^{x}\)
- The natural anti-logarithm is \(e^{x}\), often often written as \(\exp(x)\)
- Anti-log is the inverse function of log; it ‘undoes’ a log
Understanding functions in general, including \(\min(x, a)\) and \(\max(x, a)\)
Understanding indicator variables such as \([x=3]\) which can be thought of as true if \(x=3\), false otherwise, or 1 if \(x=3\), 0 otherwise
- \([x=3]\times y\) is \(y\) if \(x=3\), 0 otherwise
- \([x=3]\times[y=2] = [x=3 \,\textrm{and}\, y=2]\)
- \([x=3] + 3\times [y=2] = 4\) if \(x=3\) and \(y=2\), \(3\) if \(y=2\) and \(x\neq 3\)
- \(x\times \max(x, 0) = x^{2}[x>0]\)
- \(\max(x, 0)\) or \(w \times [x>0]\) are algebraic ways of saying to ignore something if a condition is not met
Quadratic equations
Graphing equations Once you get to multiple regression, some elements of vectors/linear algebra are helpful, for example the vector or dot product, also called the inner product:
Let \(x\) stand for a vector of quantities \(x_{1}, x_{2}, \ldots, x_{p}\) (e.g., the values of \(p\) variables for an animal such as age, blood pressure, etc.)
Let \(\beta\) stand for another vector of quantities \(\beta_{1}, \beta_{2}, \ldots, \beta_{p}\) (e.g., weights / regression coefficients / slopes)
Then \(x\beta\) is shorthand for \(\beta_{1}x_{1}+\beta_{2}x_{2} + \ldots + \beta_{p}x_{p}\)
\(x\beta\) might represent a predicted value in multiple regression, and is known then as the linear predictor