Take the set of all vectors $x = (x_1, \cdots, x_n)$ that are solutions to $p_1x_1 + \cdots + p_nx_n = I > 0$. Show that this set has $n-1$ dimensions.

I have somehow managed to get myself stuck on the last part of this proof it seems. I am not using the fact that this set is a hyperplane and that hyperplanes are $n-1$ dimensions of the space they are in.

It is easy to show that $\{x_1, \cdots x_n\}$ spans the set we are considering, since $\sum p \cdot x$ is a linear combination and all that. However, $x_n$ can be expressed as a linear combination of $\{x_1 \cdots x_{n-1}\}$:

$$x_n = \frac{I - (p_1x_1 + \cdots p_{n-1}x_{n-1})}{p_n}$$

So we can remove $x_n$ from the span and the resulting set still spans. Now we want to show $\{x_1, \cdots n_{n-1}\}$ are linearly independent. That is, if $p_1x_1 + \cdots p_{n-1}x_{n-1} = 0$, all $p_i = 0$. If the set is spanning and linearly independent, then it is a basis. Since it would have $n-1$ vectors, it would be of dimension $n-1$ and we'd be done.

So I note that $p_1x_1 + \cdots + p_{n-1}x_{n-1} = I - p_nx_n$, and that $I > 0$.

So I assume there is a case where $I - p_nx_n = 0$ and where $I - p_nx_n \neq 0$. I am not sure how to finish off this proof, which makes me sad, because I think I'm just missing something obvious. Any assistance would be appreciated.


2 Answers 2


Let matrix $A = \begin{bmatrix} p_1 & p_2 & \ldots & p_n \end{bmatrix}$. Let $\mathbf{x}^*$ be a fixed solution to $A \mathbf{x} = c$. Then for any vector $\mathbf{u}$ that belongs to the null space of $A$, we have $A \mathbf{u} = 0$ hence $\mathbf{x} = \mathbf{x}^* + \mathbf{u}$ is also a solution (furthermore, all solutions $\mathbf{x}$ can be written this way). The dimension of the set of solutions to $A\mathbf{x} = c$ is thus the dimension of the null space of $A$.

By the rank-nullity theorem, the rank of matrix $A$ plus the nullity (dimensionality of the null space) of matrix $A$ equals the number of columns of $A$:

$$ \operatorname{rank}(A) + \operatorname{nullity}(A) = n$$

Matrix A is rank 1 (assuming not all $p_i$ are zero) and hence the nullity is $n-1$.

  • $\begingroup$ I suppose yeah, this is the easy way to do it. I have never used this method before and was looking to just use linear combinations to show my desired result. I now have the answer and will post it (and in like 3 hours I will be able to upvote your answer) $\endgroup$
    – Kitsune Cavalry
    Commented Aug 17, 2017 at 20:55

I got some outside help for the ending of the proof I was attempting. I'll leave this question if by chance someone else finds it useful.

So if we want to show $p_1x_1 + \cdots + p_{n−1}x_{n−1} = 0 \implies p_i = 0 \quad \forall i$, then assume without loss of generality that $p_1 \neq 0$. We have

$$x_1 = \left(-\frac{p_2}{p_1}\right) x_2 + \cdots + \left(-\frac{p_{n-1}}{p_1}\right)x_{n-1}$$

$$\implies x_1 + \left(\frac{p_2}{p_1}\right) x_2 + + \cdots + \left(\frac{p_{n-1}}{p_1}\right)x_{n-1} = 0 \tag1$$

Then subtract $p \cdot x = 0$ from both sides.

$$(x_1 - p_1x_1) + \left(\frac{p_2}{p_1} x_2 - p_2 x_2\right) + \cdots + \left(\frac{p_{n-1}}{p_1} x_{n-1} - p_{n-1}x_{n-1} \right) = 0$$ $$\implies (1 - p_1)x_1 + \left(\frac{p_2}{p_1} - p_2 \right)x_2 + \cdots + \left(\frac{p_{n-1}}{p_1} - p_{n-1} \right)x_{n-1} = 0 \tag2$$

Set $(1) = (2)$.

$\implies 1 - p_1 = 1 \implies p_1 = 0$

which is a contradiction.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.