How to get more space in piecewise function on LaTeX?

I'm writing a paper and I have inserted a piecewise function into my tex-code. My problem is that the piecewise function is very tightly packed. How could I stretch the bracket and get more space between the formulas? See code and result below:

$$\displaystyle{ P(Y_i = j) = \left\{ \begin{array}{ll} \frac{1}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad k=K \\ \frac{e^{\beta_j\cdot x_i}}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad 1 \leq k \leq K-1 \end{array} \right., }$$

3 Answers

Don't use  for display math.

\displaystyle is not needed.

Option -1:

\documentclass{article}
\usepackage{amsmath}
\begin{document}

$P(Y_i = j) = \left\{ \begin{array}{ll} \frac{1}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad k=K \\[1em] \frac{e^{\beta_j\cdot x_i}}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad 1 \leq k \leq K-1 \end{array} \right.,$
\end{document}

Option - 2

You can use cases.

\documentclass{article}
\usepackage{amsmath}
\begin{document}

$P(Y_i = j) = \begin{cases} \frac{1}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad k=K \\[1em] %%% <--- here \frac{e^{\beta_j\cdot x_i}}{1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad 1 \leq k \leq K-1 \end{cases}$
\end{document}

In either case, you can use \$<dimen>] to change the vertical separation. Beautification To make your sums to look less cramped, you can use \limits as in \sum\limits_{k=1}^{K-1} Code: \documentclass{article} \usepackage{amsmath} \begin{document} \[ P(Y_i = j) = \begin{cases} \frac{1}{1+\sum\limits_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad k=K \\[2em] \frac{e^{\beta_j\cdot x_i}}{1+\sum\limits_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad 1 \leq k \leq K-1 \end{cases}$
\end{document}

• ehm cases ehm – yo' Dec 11 '13 at 13:16
• if you are loading amsmath anyway you could use the cases environment which is specifically for this use. – David Carlisle Dec 11 '13 at 13:16
• @tohecz I didn't finish yet. Options are on way :-) – user11232 Dec 11 '13 at 13:17
• @HarishKumar However, by that time, the OP will very likely leave, using quite not as good solution as he could. – yo' Dec 11 '13 at 13:17
• @tohecz I am slow ;-). Now I am finished. :-) – user11232 Dec 11 '13 at 13:23

The dcases environment by »mathtools« from the »mh« bundle seems to be quite handy here.

\documentclass[11pt]{article}
\usepackage[T1]{fontenc}
\usepackage{mathtools}  % loads »amsmath«

\begin{document}
$P(Y_i=j)= \begin{dcases} \frac{1}{1+\sum\limits_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad k=K \\ \frac{e^{\beta_j\cdot x_i}}{1+\sum\limits_{k=1}^{K-1}e^{\beta_k\cdot x_i}} & \quad 1 \leq k \leq K-1 \end{dcases}$
\end{document}

I had no idea that amsmath had such a useful redefinition, so that should be the correct answer (after use of cases). However, for completeness (or if you want more space) you have a few options:

• reformulate your terms to allow for more vertical leg-room
• modify \arraystretch to a desired scale (but will cause potentially awkward space on top)
• use a strut, as in the example below:

\documentclass{article}
\usepackage{array}

%\renewcommand\arraystretch{4}

\begin{document}
$P(Y_i = j) = \left\{ \begin{array}{>{\displaystyle}ll} \biggl(1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}\biggr)^{-1} & \quad k=K \\ \biggl(1+\sum_{k=1}^{K-1}e^{\beta_k\cdot x_i}\biggr)^{-1} e^{\beta_j\cdot x_i} & \quad 1 \leq k \leq K-1 \rule{0pt}{8ex} % called a 'strut' -- a vertical rule of nil width \end{array} \right.,$
\end{document}

• No, this is wrong. The parentheses are too large. Try \biggl(...\biggr) instead of \left(...\right), please. After all, why \displaystyle when `\textstyle would fit there much better? Anyways, +1 for getting rid of the fractions. – yo' Dec 11 '13 at 13:25
• @tohecz Ah, yes - \left and \right don't take the limits into account very well... And personally, while I haven't much of a leg to stand on as for conventions, if you're going to have a large cases like this, it should be \displayed, not so cramped as in \textstyle. Just personal preference probably prejudiced by handwritten math :) – Sean Allred Dec 11 '13 at 13:29