1
$\begingroup$

Let $X\neq \emptyset$ be a set, let $\text{Part}(X)$ be the set of partitions of $X$ (where we require that $\emptyset \notin P$ whenever $P\in\text{Part}(X)$).

For $P, Q\in \text{Part}(X)$ we say that $P$ covers $X$ more efficiently than $Q$ if $$\text{card}(P\setminus Q) < \text{card}(Q\setminus P), $$ and we write $P<_{\text{eff}}Q$ for this.

Is $<_{\text{eff}}$ transitive?

$\endgroup$
3
  • 1
    $\begingroup$ Since $P-Q=P-(P\cap Q)$ and $Q-P=Q-(P\cap Q)$, what you are really saying is ${\rm card}(P)<{\rm card}(Q)$ (since you are taking the same number of parts out of both partitions, namely ${\rm card}(P\cap Q)$). Or am I missing something here? $\endgroup$ Commented Dec 5, 2016 at 18:53
  • 2
    $\begingroup$ @PaceNielsen I guess it is not exactly the same thing. For example, take $X=\mathbb{N}, P=\{1,2\}, \{3\}, \{4\}, \dots$, and $Q=\{1\}, \{2\}, \{3\}, \dots$. Then $P < Q$, but $P$ and $Q$ have the same cardinality. $\endgroup$
    – Tony Huynh
    Commented Dec 5, 2016 at 21:01
  • 2
    $\begingroup$ @TonyHuynh Thanks. For some reason I was thinking finite sets. $\endgroup$ Commented Dec 5, 2016 at 21:53

1 Answer 1

3
$\begingroup$

I believe the answer is yes. Let $X\neq \emptyset$ and let $P,Q,R\in {\rm Part}(X)$.

First, note that if $|P|<|Q|$, then $P<_{\rm eff}Q$, since $|P-Q|\leq |P|<|Q|=|Q-P|$.

Now we are ready for the proof. Assume $P<_{\rm eff}Q<_{\rm eff}R$. Without loss of generality, we may remove $P\cap Q\cap R$ from all the sets, and reduce to the case that $P\cap Q\cap R=\emptyset$. By our previous paragraph, we also have $|P|\leq |Q|\leq |R|$. If inequality holds anywhere, then we have $P<_{\rm eff}R$ and we are done. Assume the contrary case holds, and so all three sets have equal cardinality. Note that $|P|$ must be infinite, else the computation from my comment above applies.

From the fact that $|P-Q|<|Q-P|$ and $|P|=|Q|$, we see that $|P\cap Q|=|P|=|Q|$. Now because $(P\cap Q)\cap R=\emptyset$, we have $|Q-R|\geq |P\cap Q|=|Q|=|R|\geq |R-Q|$, contradicting the fact that $Q<_{\rm eff}R$. Thus, this contrary case cannot actually happen after all.

$\endgroup$
1
  • $\begingroup$ Great answer and easy to understand -- thanks Pace! $\endgroup$ Commented Dec 6, 2016 at 7:14

You must log in to answer this question.

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