Reader Comments

Post a new comment on this article

inference of causal relationships

Posted by Debora on 05 Mar 2013 at 07:47 GMT

To me it is not clear that one can infer cause based on the regression of breeding time against nesting mortality. The question here is whether annual daily nesting mortality rate (DMR) causes a change in breeding phenology, or whether it is the change in breeding phenology that leads to a change in experienced DMR. I do not doubt that patterns of nesting mortality can be important for phenology, but given the evidence/analyses presented for this specific study I am doubtful about the main conclusion claiming a causal link from DMR to breeding phenology .

No competing interests declared.

RE: inference of causal relationships

janhus replied to Debora on 06 Mar 2013 at 13:03 GMT

Dear Debora! Thank you very much for your comment!

We agree that inferring causality from regression is in general tricky. Experimental manipulation is often the only way to reveal causal relationships.
Our study is in principle correlative. As you say, it can therefore be claimed about one of our results that either 1) higher nesting mortality shifts distribution of breeding times to later dates, or 2) population of shrikes breeding later suffers on average higher nesting mortality.

Ad 1) Shrikes rarely establish second clutch. Most typically, replacement clutch is established only if the first clutch fails. As the proportion of failed first clutches increases (when the nesting mortality is high), more replacement clutches are established. Consequently, overall annual distribution of breeding dates is shifted towards later calendar dates. Based on this prior knowledge of the breeding biology it is possible to indirectly infer causality. Higher nesting mortality always predicts shift in the overall distribution of phenology to later calendar dates.

Ad 2) Although claiming that the later breeding causes higher nesting mortality is in theory also possible, more assumptions would have to be made in order to infer causality. First, one has to assume that nesting mortality increases during the breeding season. Second, this pattern has to be consistent between years. Solid prior knowledge backuping these underlying assumptions is however missing. Nesting mortality may both increase and decrease during the season (see figure 1d). Breeding phenology therefore does not predict seasonality in nesting mortality.
In shrikes, it is unlikely that later breeding exposes nests to new predators. Relevant nest predators of shrikes are
present and active all around the year. Finally, is is also important to note here that seasonality in DMR is related to exposure date of the nest, not to the date of onset of breeding.

Jan Husek on behalf of all the authors

No competing interests declared.