Tools
Tools
Boakes, Zach and Stafford, Rick (2024) Figure S.5. 3: R code and priors files used to create the model.
Bayesian Belief Network R code and priors files
| Alternative Title: | Is tourism helpful or harmful for coral reef health? Stakeholder assessment, actions and management measures in three reef-based tourism areas in Bali Indonesia. | ||||||
|---|---|---|---|---|---|---|---|
| Research / Data Type: | Model | ||||||
| Creators: |
|
||||||
| Groups: | Faculty of Science & Technology | ||||||
| Collection period: |
|
||||||
| Date: | January 2024 | ||||||
| Date type: | Completion | ||||||
| Data collection method: | We quantified the findings of our qualitative and ecological data collection using a Bayesian belief network (described in full in Stafford et al. (2020)), allowing us to predict the outcomes of 10 key scenarios for reef-based tourism areas for coral reef health. This network was made up of a series of ‘nodes’ which were connected by weighted ‘edges’ (circles and connecting lines, respectively). The weights of each edge was based on changes likely to occur to receiving nodes (known as ‘child’ nodes) given a change in the ‘parent’ or originating node. These positive interactions (if the parent node increases, it is most likely that the child node will also increase) or negative interactions (if parent node increases, the child node will most likely decrease), the values give to the edges were determined from combinations of previously published literature, results of the qualitative study and expert opinion. The edge values contributing most to overall variation in fish and coral health values were determined and validated through a sensitivity analysis approach. Certain nodes were given ‘prior’ values, ranging from -4 to 4. which progress through the network, with child nodes becoming parent nodes for subsequent interactions. The priors changed depending on the given scenarios investigated, which were highlighted as plausible actions stakeholders may take to protect coral reefs in reef-based tourism areas. To help quantify uncertainty in the network, each scenario was run 10,000 times. The first run uses the exact model as provided (and forms the circular point in the results figures), the remaining runs involve randomly selecting 10 % of interactions in each run and adjusting each of them by a randomly determined amount of up to ± 0.8. 95% confidence intervals of the output of each parameter are calculated by removing the highest and lowest 2.5 % of values. The ten scenarios were given in table 2. The full mathematics of Bayesian belief networks which were used in this study are explained in depth in Stafford et al. (2020). | ||||||
| Keywords: | Bayesian Belief Network Model | ||||||
| Status: | Unpublished | ||||||
| Contact email address: | bordar@bournemouth.ac.uk |
| DOI: | https://doi.org/10.18746/bmth.data.00000340 |
|---|---|
| Depositing User: | Zach Boakes |
| Year Deposited: | 17 Jan 2024 16:29 |
| Revision: | 16 |
| Last Modified: | 17 Jan 2024 16:30 |
Available Files
Full Archive
|
Administer Item