Setting time intervals between captures

[jjroper]

Dear Forum,

I am just here to make an observation and perhaps to verify that I am thinking correctly. When setting up a new model, using recaptures, one of the options is to set time intervals between capture sessions. An assumption is that capture sessions occur at uniform time intervals, and when not, if they are very variable, then this allows one to control for this variability. I just happen to have some frog capture data that happened at variable intervals.

If one does not use this option, then the time interval involved is whatever time interval there is between capture sessions. So, if one captured biweekly, then the survival rates calculated will be per biweek. Okay, let's imagine that the time interval averages 2 weeks between captures, but for climatic reasons, can vary from 5 to 25 days. Now, we could put fractions of two weeks into those intervals, or we could put days into those intervals.

Now, if we use days, then our survival rates will be per day, rather than per capture interval, while if we use the fraction of the capture interval, then the survival rate will be in capture intervals. Following this train of thought, if age were important and we used a model to examine age, our Phi will be in age units - that is, if survival is over a maximum of 5 sampling sessions, then we will get five estimates for Phi (one for each age), however, if we used DAYS to set the interval between capture sessions, the value will be in daily survival rates. If we used fractions of time of two weeks (the average interval between capture sessions), then the survival rates will be for two-week intervals.

Thus, one must have some preferred notion of what time interval is of interest, because we can get MARK to correctly estimate the confidence interval. But, if we have daily survival rate and we wish to actually use monthly survival rate, then we will have to recalculate by hand the new confidence intervals (if I speak correctly).

Anyone want to comment or confirm (or clarify ) these observations?

Cheers,

Jim

[Eurycea]

Jim, it makes sense to me, but take that with a grain of salt...

If you have daily intervals but want a monthly survival rate, it is perhaps just easiest to tell Mark what your intervals are in terms of months.

[cooch]

Dear Forum,

I am just here to make an observation and perhaps to verify that I am thinking correctly. When setting up a new model, using recaptures, one of the options is to set time intervals between capture sessions. An assumption is that capture sessions occur at uniform time intervals, and when not, if they are very variable, then this allows one to control for this variability. I just happen to have some frog capture data that happened at variable intervals.

If one does not use this option, then the time interval involved is whatever time interval there is between capture sessions. So, if one captured biweekly, then the survival rates calculated will be per biweek. Okay, let's imagine that the time interval averages 2 weeks between captures, but for climatic reasons, can vary from 5 to 25 days. Now, we could put fractions of two weeks into those intervals, or we could put days into those intervals.

Now, if we use days, then our survival rates will be per day, rather than per capture interval, while if we use the fraction of the capture interval, then the survival rate will be in capture intervals. Following this train of thought, if age were important and we used a model to examine age, our Phi will be in age units - that is, if survival is over a maximum of 5 sampling sessions, then we will get five estimates for Phi (one for each age), however, if we used DAYS to set the interval between capture sessions, the value will be in daily survival rates. If we used fractions of time of two weeks (the average interval between capture sessions), then the survival rates will be for two-week intervals.

Thus, one must have some preferred notion of what time interval is of interest, because we can get MARK to correctly estimate the confidence interval. But, if we have daily survival rate and we wish to actually use monthly survival rate, then we will have to recalculate by hand the new confidence intervals (if I speak correctly).

Anyone want to comment or confirm (or clarify ) these observations?

Cheers,

Jim

The mechanics/idea behind how MARK handles unequal intervals are discussed in some detail in chapter 4 (sidebar beginning on p. 26, continuing over next 3-4 pages). As noted on p. 28, a key part of this is 'identifying the appropriate interval' (which should be motivated by biology), and then getting MARK to allow you to make inference for that interval, irrespective of what the sampling interval was when the data were collected.