Is this a valid proof using induction?

80 Views Asked by At

I wanna show the binomial distribution is a valid distribution that is:

given $$p_k= {n\choose{k}}(1-p)^{n-k}p^k$$ prove

$$\sum_{k=0}^{n} p_k=1$$

I did it using induction.

First the base case is $n=1$ in which $$\sum_{k=0}^{1} p_k=1$$

Then the inductive step, let $$\sum_{k=0}^{n}{n\choose{k}}(1-p)^{n-k}p^k=1$$ hold for specific $n$ show that $$\sum_{k=0}^{n+1} {n+1\choose{k}}(1-p)^{n+1-k}p^k=1$$

What I did is that I made a change of variables so that let $n+1=m$, therefore $$\sum_{k=0}^{n+1} {n+1\choose{k}}(1-p)^{n+1-k}p^k=\sum_{k=0}^{m} {m\choose{k}}(1-p)^{m-k}p^k=1$$

This proof seems too fishy for me. So can someone tell me if it's valid or not, and if not why?

2

There are 2 best solutions below

0
On BEST ANSWER

You've made one of the common mistakes in induction. What often happens is that people start with some messy formula involving $n$ and then say "$n$ is just a variable, so I can substitute $n+1$ for $n$ and then I have some messy formula involving $n+1$ and I'm done!"

The way that I like to avoid this error, is that you make your inductive hypothesis different. You assume that the claim is true when $n=m$ and use that statement to prove the statement of interest in the case when $n=m+1$. In this set-up, you must think of $m$ as fixed. This prevents you from confusing when $n$ is variable and when $n$ is fixed.

In this problem, Pascal's Identity might be helpful (along with separating a sum, factoring, and a change of variables so that both sums have the same limits).

0
On

In addition to Michael Burr's comments, you should also show how you proved $$\sum_{k=0}^{1} p_k = 1$$ (rather than basically just restating it)