FAQs

Frequently Asked Questions

Below we provide a list of the questions most common among climwin users. This list is by no means exhaustive. For any further questions you might have please visit our forum and ask the climwin community.

Q: What statistical models are compatible with climwin?

A:

climwin is designed to run with a variety of data structures and model functions available in R. Currently, we support both general and generalised linear models (e.g., gaussian, poisson, binomial) and functionality for mixed effects models through both the lme4 and nlme packages. Users can also employ proportional hazard models using the coxph function from the survival package.

Q: Can I run climwin with weekly/monthly data?

A:

Climate window analyses will ideally use high resolution daily climate data; however, in reality, we appreciate that this scenario is not always possible. The newest version of climwin allows for the use of climate data collected at weekly or monthly intervals. This will require the user to specify the 'cinterval' argument in climwin as "week" or "month" where appropriate.

Q: Should I use the metric "C" or "AIC" for calculating p-values?

A:

The pvalue function provided in climwin determines the likelihood that a given climate window result represents a false positive. Therefore, to calculate our p-value we must first estimate the distribution of deltaAICc values we would expect to obtain by chance, using the function randwin. Ideally, users will run a large number of iterations with randwin; however, it may often take a long time to run a single climwin analysis and large numbers of replication may be unrealistic. Metric "C" is suited for pvalue calculations when few iterations have been conducted (<100), while metric "AIC" is suitable for cases with a large number of iterations.

NOTE: The methods employed by climwin are exploratory in nature, which will enhance the possibility of detecting false positives. All climwin results should be checked using the pvalue function.

Q: How do I deal with missing climate data?

A:

There are a number of potential methods for dealing with missing data, some of which are now incorporated into climwin:

• Estimation from nearby climate records. With the argument cmissing = "method1" in the slidingwin\randwin\singlewin functions climwin will replace any NA values with the mean climate in the two climate records preceding and following the missing record. NOTE: This method will not work when large blocks of consecutive NA values are present (i.e. whole months are missing).

• Estimation from alternative years. With the argument cmissing = "method2" in the slidingwin\randwin\singlewin functions climwin will replace any NA values with the mean climate from all records on the same date.

• Reduce resolution of analysis. By setting cinterval to either week or month climwin will estimate the mean climate for every week/month respectively, excluding any NA records.

• Estimation from alternative weather stations. In field sites with multiple available weather stations NA records that exist in one weather station can be replaced by climate data from alternative stations in the region. This will require users to replace NA values before climwin analysis.