# Center the terms in an equation?

I have the following equation:

Notice how the last two terms are not centered? If I use the code specified below, I would get that effect. Now, what I would like to have is the effect shown in the first term (centered the term "Bias"). You probably also see that since I kind of fail with paint, the first terms are not centered correctly.

Now I was wondering if it is possible to do that in Latex? Notice that it should be centered between the equal sign and the plus sign.

In code form:

\begin{aligned} Err(x_{0}) &=\left(E\left[\hat{f}(x_{0})\right]-f(x_{0})\right)^{2}&+&E\left[\hat{f}(x_{0})-E\left[\hat{f}(x_{0})\right]\right]^{2}&+&\sigma^{2}_{\epsilon} \\ &=\text{Bias}^{2}&+&\text{Var}(\hat{f}(x_{0})&+&\text{Var}(\epsilon) \end{aligned}


Now unfortunately, the terms in the first equation and the second equations are not centered. I was wondering how I can center each term.

Thanks

• $$...$$ is outdated. Please use $...$ instead. Your first example is also in code form ;-) And provide a MWE, please
– user31729
Feb 18 '15 at 15:52
• Plus please learn the correct use of the \text command, hint: you are not using it right. Define \Bias, \Var in the preamble using \DeclareMathOperator\Bias{Bias}, and \DeclareMathOperator\Var{Var}. Var and Bias should be upright in all cases, \text is not! Feb 18 '15 at 15:59
• if you want each line centered separately, use gathered instead of aligned. Feb 18 '15 at 16:20
• You shouldn't use \left and \right to begin with. Feb 18 '15 at 21:20

I wouldn't center the terms, but it's just my opinion.

You can obtain correct spacing around the = and + symbols with alignedat or also with array if you prefer centering.

\documentclass{article}
\usepackage{amsmath,array}

\DeclareMathOperator{\Err}{Err}
\DeclareMathOperator{\Bias}{Bias}
\DeclareMathOperator{\Var}{Var}
\DeclareMathOperator{\E}{E}

\newcolumntype{O}{>{{}}c<{{}}}

\begin{document}

\begin{alignedat}{3} \Err(x_{0}) &= \bigl(\E[\hat{f}(x_{0})]-f(x_{0})\bigr)^{2} &&+ \E\bigl[\hat{f}(x_{0})-\E[\hat{f}(x_{0})]\bigr]^{2} &&+ \sigma^{2}_{\epsilon} \\ &= \Bias^{2} &&+ \Var(\hat{f}(x_{0})) &&+ \Var(\epsilon) \end{alignedat}

$$\setlength{\arraycolsep}{0pt} \begin{array}{r O c O c O c O c} \Err(x_{0}) &=& \bigl(\E[\hat{f}(x_{0})]-f(x_{0})\bigr)^{2} &+& \E\bigl[\hat{f}(x_{0})-\E[\hat{f}(x_{0})]\bigr]^{2} &+& \sigma^{2}_{\epsilon} \\[1ex] &=& \Bias^{2} &+& \Var(\hat{f}(x_{0})) &+& \Var(\epsilon) \end{array}$$

\end{document}


Note the definitions of the four math operators; the \newcolumntype declaration is for the second rendering, it represents the column where the = and + signs are.

Don't use \left and \right, that produce awfully oversized fences. Just \bigl and \bigr are necessary in two places.

In general you should not use align.. for that kind of construction. If the equations are not numbered a solution is to use array

\documentclass{article}
\usepackage{amsmath}

\DeclareMathOperator{\Bias}{Bias}
\DeclareMathOperator{\Var}{Var}

\begin{document}

\begin{equation*}
\setlength\arraycolsep{1.5pt}
\begin{array}{rcccccc}
Err(x_{0}) & = & \left(E\left[\hat{f}(x_{0})\right]-f(x_{0})\right)^{2}
& + & E\left[\hat{f}(x_{0})-E\left[\hat{f}(x_{0})\right]\right]^{2}
& + & \sigma^{2}_{\epsilon}\\
& = & \Bias^2
& + & \Var(\hat{f}(x_{0})
& + & \Var(\epsilon)
\end{array}
\end{equation*}

\end{document}


producing

See this answer for the rationale for setting the array inter column space to 1.5pt.