I am trying to understand the article "A note on the union-closed sets conjecture" by Roberts and Simpson.
At the first paragraph of the second page they write:
"We say that $A$ is a basis set in $\mathcal{A}$ if it is not the union of other sets in $\mathcal{A}$. Suppose that $\mathcal{A}$ is a counterexample with $\mid$$\mathcal{A}$$\mid = 2n + 2$ for some integer $n$. Then no element occurs in more than $n$ of its sets. If we remove a basis set from $\mathcal{A}$, then we get a smaller collection which is also a counterexample. A counterexample of minimum size must therefore contain an odd number of sets."
- How to show that $\mathcal{A}$ has always a basis set?
- They say that removing a basis set from a family with an even number of sets gives a smaller collection which is also a counterexample: I don't understand why this does not apply starting from a family with an odd number of sets.
If $\mathcal A$ is finite collection of sets then it has natural partial order using subsets, and so one or more smallest non-empty members of $\mathcal A$. Each of these is a basis set
With $|\mathcal A| =2n+2$ as a counter example, each element would appear no more than $n$ times. Remove a basis set to give $\mathcal A^\prime$, and you would have $|\mathcal A^\prime| =2n+1$ and each element would appear no more than $n$ times, so $\mathcal A^\prime$ would be a smaller counter example
This argument does not work for odd-sized sets: with $|\mathcal B| =2n+1$ as a counter example, each element would appear no more than $n$ times. Remove a basis set to give $\mathcal B^\prime$, and you would have $|\mathcal B^\prime| =2n$ and each element would appear no more than $n$ times, so you could not show whether $\mathcal B^\prime$ would be a smaller counter example