In order to answer the op's question, is it better to breed a parent to a child or a sibling to a sibling, it is important to understand the probability of inbreeding depression. In other words, how likely is it that inbreeding will weaken the next generation in a measurable way.


To make it easy we will envision an organism with just one gene pair. This one gene has 4 genotypes A,B, a and b. The mother is AB and the father is ab. Using a punnet square, the offspring will be in a ratio of. Aa, Ba, Ab, Bb. Now let us note that BB is a homozygous negative trait. Any organism with BB will suffer a dramatically shorter lifespan. If we breed two siblings together, we could get many different results, but we should notice that only a very few pairings could produce BB. In fact only Ba x Bb could result in a BB genotype. We could work out all of the combinations or you can just trust me when I tell you there are 16 possible combinations and 2 of them result in inbreeding depression. We can reduce the ratio to 1 in eight sibling combinations can cause deleterious effects.


If we breed offspring to parents in this case, the mother can produce BB in about 1 out of two pairings. The father can never produce BB. So in this case if the mother and father reproduced equally with siblings we would find that one quarter of the pairings would able to produce BB genotypes. In this case it would be better to pair siblings with siblings than parents and offspring, but it would be better to cross offspring with the father than siblings with siblings. Even in this very simple case, the answer is ambiguous at best.


Here is the kicker, the numbers work out differently for every possible combination of traits. If the parents are Ab x Ab then BB can never happen and all pairings are safe. Try figuring out the other possible crosses by yourself. Use punnett squares. It isn't hard, but it takes a while.


http://en.wikipedia.org/wiki/Punnett_square


Now remember, there is no such thing as an organism with just one gene that has four genotypes. In reality the fitness of organsims is incredibly complex and the actual inbreeding depression in previously non inbred lines ends up being pretty low.


So in some cases it is better to breed siblings, but in reality, unless this math is done for every gene in the entire organsim, it is really difficult to tell which is best, and any difference is likely to be so small it doesn't really matter in the real world. The problems occur with sustained inbreeding over long periods, when deleterious alleles start to become concentrated. This can be especially dangerous if a desired trait is linked to a deleterious trait. Line breeding for the positive trait can quickly produce very unsound animals.


The ultimate answer to the OP's question is, well it depends, and it is probably such a small difference that it isn't important. Limit line breeding whenever possible, and out breed as much as possible without loosing the selected trait. As long as you keep the inbreeding to a minimum, the possibility of running into a problem is not high, and culling can be used to offset those problems.


David