18
$\begingroup$

Prove that $$\sum^{n}_{i=1}\binom{n}{i}i=n2^{n-1}$$

I can't find counting interpretations for either of the sides. A hint of "if $S$ is a subset of $\{1, . . . , n\}$ and $S^\prime$ is its complement then $|S| + |S^\prime| = n$" was also given, but I still don't know how to begin.

$\endgroup$

6 Answers 6

44
$\begingroup$

We can interpret this combinatorially as the number of ways to form a committee (of any size) with one chairman out of a group of $n$ people.

From $n$ people we first pick a committee of size $i$, then choose one the $i$ committee members to be the chairman. There are ${n \choose i}$ options for the members of the committee, after which there are $i$ options for the chairman. If we sum over all $i$ from $1$ to $n$, that covers committees of all possible (nonzero) sizes. So, we have $\sum_{i=1}^n {n \choose i}i$.

On the other hand, we could first pick one person from the $n$ people to be the chairman. Then for each of the remaining unchosen $n-1$ people, they can be either in or out of the committee. That's $2^{n-1}$ possibilities. So, we have $n2^{n-1}$.

$\endgroup$
2
  • $\begingroup$ I really like this "committee" proof. Do you have any reference where such methods are used to prove more complicated sums? $\endgroup$
    – Prism
    Commented Jun 2, 2013 at 1:36
  • 2
    $\begingroup$ @Prism The technique is called "double counting." Proofs from the Book by Ziegler has a chapter on cool double counting proofs, which has a partial preview here. $\endgroup$
    – Alexander Gruber
    Commented Jun 2, 2013 at 2:47
17
$\begingroup$

Hint: consider the the set of all subsets of $\{1,2,\dots,n\}$ (of which there are $2^n$) and try to find the total sum of the sizes of the subsets in two different ways. For example, the possible subsets of $\{1,2\}$ are $\{\},\{1\},\{2\},\{1,2\}$. Then adding up the sizes of each subset gives $0+1+1+2 = 4$.

More explicitly, if we add up the sizes of all possible subsets of $[n]=\{1,2,\dots,n\}$, we can either:

$1)$ Note that there are $\binom{n}{i}$ subsets of size $i$ which gives that the total sum of sizes is

$$\sum_{i=1}^{n}\binom{n}{i}i$$

$2)$ Observe that each element is in $2^{n-1}$ subsets, and so contributes $2^{n-1}$ to the total sum. Thus the sum equals

$$n2^{n-1}$$

The value of the sum doesn't change regardless of how we do it, so the expressions must be the same.

$\endgroup$
3
  • 1
    $\begingroup$ Could you explain more why each element is in $2^{n-1}$ subsets? $\endgroup$
    – ithisa
    Commented May 13, 2013 at 22:27
  • $\begingroup$ @Eric Certainly. I have two reasons, pick your favourite. $1)$ Each subset that contains a given element may or may not contain each of the other $n-1$ elements, so there are $2^{n-1}$ possible such sets, one "$2$" for each choice. (This approach is the same as one of the ways to prove that the number of subsets of $[n]$ is $2^n$.) $2)$ We can partition the set of subsets of $[n]$ into the sets that contain the given element and the sets that don't. There is an obvious bijection between these two partitions, so they must be the same size, $2^n/2=2^{n-1}$. (This is my preferred method.) $\endgroup$ Commented May 13, 2013 at 22:41
  • $\begingroup$ So its basically sum of lengths of all possible subsets... $\endgroup$
    – Mahesha999
    Commented Mar 10, 2017 at 4:56
6
$\begingroup$

This is not really an answer to the question (very good ones have already been given), but to the more daunting challenge of finding a way to actually use the hint given in the question (in an interesting way).

The left hand side $L=\sum_{i=0}^n\binom nii$ gives the sum of the sizes of all subsets of an $n$-element set, grouping together the $\binom ni$ subsets of size$~i$. Note that I've thrown in the empty set, which makes no difference for this sum, but makes the number of subsets summed over equal to $2^n$. Since taking the complement of all subsets again gives every subset exactly once (the operation is an involution), $L$ also gives the sum of the sizes of the complements of all subsets of an $n$-element set. But if a set has size $i$, then its complement has size $n-i$ (this is where the hint is used!) so this means that $L=\sum_{i=0}^n\binom ni(n-i)$. Adding up the two summations gives $$ 2L=\sum_{i=0}^n\binom ni(n+(n-i))=n\sum_{i=0}^n\binom ni=n2^n. $$ Dividing both sides by$~2$ gives the desired equation $\sum_{i=0}^n\binom nii=L=n2^{n-1}$.

$\endgroup$
1
$\begingroup$

Another possible interpretation is through the expected number of heads in $n$ tosses of a fair coin which is $n/2$.
That is, we shall have $$ {n \over 2} = \sum\limits_{\left( {0\, \le } \right)\,i\,\left( { \le \,n} \right)} {i\left( \matrix{ n \cr i \cr} \right)\left( {{1 \over 2}} \right)^{\,i} \left( {{1 \over 2}} \right)^{\,n - i} } = {1 \over {2^{\,n} }}\sum\limits_{\left( {0\, \le } \right)\,i\,\left( { \le \,n} \right)} {i\left( \matrix{ n \cr i \cr} \right)} $$

$\endgroup$
1
$\begingroup$

Assume :

You have $n$ beads, numbered $1$ to $n$, of same mass $=1$.

1)$\binom{n}{i}$ is the number of distinct sets of $i$ beads, where $i=0,1,2,...n.$

Every set of $i$ beads has a mass $i\cdot1$, there are $\binom{n}{i}$ of them, hence their total mass:

$ T:= \sum_{i=0}^{n}\binom{n}{i}i\cdot 1.$

2)Separate bead $1$ and consider the total number of sets that can be formed with the remaining $n-1$ beads: $2^{n-1}$.

Hence adding bead $1$ to each of the above sets we conclude:

Bead $1$ is present in $2^{n-1}$ sets out of the $2^n$ original sets.

Bead $1$ contributes $2^{n-1} \cdot 1$ to $T$, so do the beads $2,3,...n $.

Hence : $T=n2^{n-1}\cdot 1$.

$\endgroup$
1
$\begingroup$

Consider an ordered $n+1-tuple$ of numbers,assigned to each subset of the set $S=\{1,2,3,...n,n+1\}$, whose $i$th coordinate is either $1$ or $0$ accordingly as the the $i$th object is chosen or not respectively. Clearly the left hand side of the equation gives the total number of $1$'s on all the possible $n+1-tuple$s. Again for every $k$ belonging to S a total of $2^n$ $1$s are contributed in those $n+1-tuple$s for $k$.Therefore total no. of $1$s contributed for all $n+1$ numbers is $(n+1)2^n$.Therefore $\mathrm{LHS=RHS}$ Proven.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .