Background | Parameters included | Results | Figure | Table 1 — Reference Papers | Table 2 — Parameters | Table 3 — Risk Factors | Methods | Appendix: Code to Reproduce the Analysis | Data loading | Data processing | Summary table | Parameter & Risk factor tables | Plot generation

Data model diagramme | Areas data | Population data | Survey data | Programme data

0. Prepare webtool GeoJSON input | 1. (Up)Load data inputs | 2. Choose model areas and time points | 3. Review input data | 4. Prepare model inputs

E101 | E102 | E103 | E104 | E105 | E106 | E107 | E108 | E109 | E110 | E111 | E112 | E113 | E114 | E115 | E116 | E201 | E202 | E203 | E204 | E205 | E206 | E207 | E208 | E209 | E210 | E211 | E212 | E213 | E214 | E215 | E216 | E217 | E218 | E219 | E220 | E221 | E222 | E223 | E224 | E225

Background | Why do we need backimputation? | A simple toy scenario | Real world scenario: UK COVID deaths | No left-censoring: early R~t~ estimates remain similar | Left-censoring: back-imputation reduces bias. | Caveats | References

Input data | Step A: Serial Interval | Step B: Incidence | Estimate R~t~ | Step C: estimate_R() | Prior for R~t~ | Forecast future incidence | Step D: Projections package | Example A: Entire workflow | Example B: Estimating R~t~ using non-parametric distribution | Example C: Estimating R~t~ using infector-infected cases | Example D: Estimating R~t~ using "si_from_sample" | Example E: Estimating R~t~ using "uncertain_si" | Example F: Changing the prior for estimating R~t~ | Example G: Changing the time window to estimate R~t~ | References

SARS-CoV-2 variants

Estimate R~t~ from temporally aggregated incidence data | estimate_R() for aggregated data | Estimate R~t~ from weekly COVID-19 data | References

Overview | Estimating R on sliding weekly windows, with a parametric serial interval | Estimating R with a non parametric serial interval distribution | Estimating R accounting for uncertainty on the serial interval distribution | Estimating R and the serial interval using data on pairs infector/infected | Changing the time windows for estimation | Accounting for missed generations of infections | Different ways of specifying the incidence | Specifying imported cases | EpiEstimApp

Using the P. vivax model | Model details | Parameters | Structure | New Infections | Immunity | Infectivity of LM-detectable infections | Hypnozoites | Drug treatment | Equilibrium | Key Model References

Estimating variation in parameters

Variation and population size | Parameterisation | Simulations | Stochastic elimination | Estimating variation

Setting bed net parameters | Parameterisation | A note on mosquito species | Simulation | Visualisation | Comparing coverage and population bed net usage | Using the netz package

General syntax | Variables | Continuous time models | Basic operations | Conditionals | Operators | Mathematical functions and constants | Arrays | Array size | Special functions for arrays | Distribution functions | Semantics of random number draws | Special functions | Parameters | Data | Interpolation | Restricted names

Background | Data inputs | Tool workflow | SHIPP Tool estimates process | Limitations of tool | Usage of tool outputs

Order of events | Options | odin2.compatibility | odin2.check_bounds | odin2.target

Rebuilding the bootstrap library | Testing a copy of hipercow on the cluster | Recreating the vignettes | Rtools, Java support and R versions (Windows) | Adding R Versions | Updating RTools | Updating Java support | stan

Using cyphr

Type of location | An example | Possible working patterns | Pulling complete trees, or not | Moving as little data as possible | Interaction between source tree and .outpack | Sharing packets with collaborators using a shared file system

The basic idea | An example | Create a tiny package | Handle the configuration | Evaluate the query | Trying it out | Making the plugin more robust | Saving metadata about what the plugin did | Potential uses

Using dependencies | Basic use | Filtering candidates by parameters | Interpreting errors | Filtering candidates in other ways | Computing dependencies and using many dependencies at once | How dependencies interact with locations | Other points

orderly.quiet | orderly.index_progress | orderly.schema_validate | orderly.disable_orderly2_compat | orderly.git_error_ignore | orderly.git_error_is_warning | orderly.interactive_parameters_missing_error | Environment variables

Installation | Creating an empty orderly repository | Creating your first orderly report | Depending on packets from another report | Available in-report orderly commands | Parameterised reports | Shared resources | Strict mode | Interactive development | Deleting things from the archive | Interaction with version control | Interaction with the outpack store | Relationship between orderly and outpack

Summary of changes | So long YAML and thanks for all the whitespace errors | Implications | Database support has been moved into a plugin | No more commit | No more testing or development mode | New, language-agnostic, backend | Other, smaller changes | What is missing compared with orderly1 | How to migrate | What about the packages? | What were the problems in version 1 | Working with old versions without migrating | Users of orderly2 | Migrate your source code | Use the built-in compatibility support | Install the orderly2 compatibility package

Collaborative analysis | Reproducible analyses | Difference from a workflow system | What is reproducibility anyway? | I'm in, where do I start?

Structure of queries | Special simple queries | Scoping queries | Scoping on name | Dependencies | Possible future queries and interface improvements | Simple things | Explain the query

Basic overview | What goes into a packet | Types of users of outpack | Directory layout | Configuration (.outpack/config.json) | Packet metadata (.outpack/metadata/) | Location information (.outpack/location/) | A file store (.outpack/files) | An archive (archive/ by default) | Adding a packet | Adding metadata | Adding files | Marking the packet as unpacked and known locally | Ordering of operations | Details | Outpack ids | Location ids | Representing hashes | Times | File store

Outpack files accidentally committed to git | I don't care about my git history at all | I just want this to go away and nothing I have committed is very large | I care about my history but want this stuff gone

Configuration | Use within a report | More on configuration | Environment variables for passwords | Advanced database configuration | Migrating from orderly v1 | Other differences in behaviour

Discrete time stochastic SIR model

E0001 | E1001 | E1002 | E1003 | E1004 | E1005 | E1006 | E1007 | E1008 | E1009 | E1010 | E1012 | E1013 | E1014 | E1015 | E1016 | E1017 | E1018 | E1019 | E1020 | E1021 | E1022 | E1023 | E1024 | E1025 | E1026 | E1027 | E1028 | E1029 | E1030 | E1031 | E1032 | E1033 | E1035 | E1036 | E1037 | E1038 | E1039 | E1040 | E1041 | E1042 | E1043 | E1044 | E1045 | E1046 | E1047 | E1048 | E1049 | E1050 | E1051 | E1052 | E1053 | E1054 | E1055 | E1056 | E1057 | E1058 | E1059 | E1060 | E1061 | E1062 | E1063 | E1064 | E1065 | E1066 | E1067 | E1068 | E1069 | E1070 | E2001 | E2002 | E2003 | E2004 | E2005 | E2006 | E2007 | E2008 | E2009 | E2010 | E2011 | E2012 | E2013 | E2014 | E2015 | E2016 | E2017 | E2018 | E2019 | E2020 | E2021 | E2023 | E2024 | E2025 | E2026 | E2027 | E2028 | E2029 | E3001 | E3002 | E3003

When might you want to create a new sampler? | A toy problem | A toy sampler

Issues | Git | Code organisation | Pull Requests | Microbenchmarks | Wishlist

Introduction | The problem | The process | Example | Running the report | Creating a report | Resources, sources and artefacts | Using artefacts from other reports | Parameterised reports | Using global resources | Using version control | Using SQL databases | Configuration | Use within a report | Advanced use | Customising the database configuration | Advanced database configuration | Using environment variables and secrets | Developing a report

Overview | The process | Limitations

The orderly changelog | User's perspective | Details

Individual researcher | Why not commit the archive to git? | Small group, sharing via shared disk | Some comments on collaboration | Group of researchers, sharing via server | Using staging environments | Dependency resolution | Looking forward, a fully decentralised approach

OrderlyWeb | Configuring orderly to talk to OrderlyWeb | Interacting with the remote server | Future developments | Deploying a remote server

Setting the scene | Comparing to data | Stochastic likelihood with a particle filter | Inference with particle MCMC (pMCMC) | Effective sampling | Comparing against vectors of data | Deterministic models from stochastic | Differentiable models | Running multiple parameter sets at once | For dust2 systems | For dust2 filters/unfilters | For monty models | Further reading

New features | Planned | Hoped | Missing features | Changes in syntax | user() becomes parameter() | Compare keyword is now removed. | Vector parameters assign without array indices | Vector/matrix/array parameters whose size is determined by input require rank argument | Interpolate results assign without array indices | Discrete-time models have a more solid time basis | Assignments to dt | Assignments to time | Use of step | The name of the time variable in continuous time models has changed | Random number function calls have changed | System size cannot be changed after creation | Changes in the way arrays are handled | Changes to how delays are supported | delay() cannot be used in discrete time models | delay() results assign without array indices | The first argument to delay() must be the name of an equation | The 'default' argument to delay() has been removed | The delay time must be constant | General changes | Known limitations | Much slower compilation time | Loss of features from odin 1.x | Updating old code | Avoiding output() in discrete-time systems

First steps, a random walk | Support code | Annotations | Type definitions | The methods | Multiple variables, the SIR revisited | Continuous time (ODE) models | Continuous time models with additional output | Continuous time models with special variables

Options | Influential options | hipercow.auto_install_missing_packages | hipercow.progress | hipercow.timeout | hipercow.validate_globals | hipercow.max_size_local | hipercow.default_envvars | hipercow.development | DIDE options | rrq options | hipercow.rrq_offload_threshold_size | Options from other packages | Setting options | R versions | Workflow considerations | Long running tasks | Disk space

Pre-requisites:- | Authentication with DIDE | About our usernames and passwords | Networks | Filesystems and paths | Cluster-based storage and home directories | Mapping network drives on your computer | Windows | macOS | Linux | Initialisation | Does it work? | Default environment variables | The Nodes | Cluster Storage | Cluster Storage on Windows Compute nodes | Cluster Storage on Linux Compute nodes

Basics | Loading packages | Loading your own functions | Defining globals | Other points | Relationship with provisioning

Clusters and Platforms | Installing prerequisites | Filesystems and paths | Initialising | Running your first task | Using functions you have written | Getting information about tasks | Fetching information with task_info() | Fetching logs with task_log_show | Watching logs with task_log_watch | Running many tasks at once | Bulk call, or "parallel map" | Bulk expression | More on bundles | Picking bundles back up again later | Making bundles from tasks | Parallel tasks | Understanding where variables come from | Cancelling tasks | Retrying tasks

Differences | Mapping of didehpc methods to hipercow functions | queue_didehpc | task | bundle | Configuration | Missing features | New features

Using pkgdepends | A list of packages | Manually | Automatically, from an environment | Using a script | Using renv | Some details about the process

Task-level parallelism | What resources does the cluster have? | Specifying multi-core resources | Running parallel tasks | Using the Parallel package | Using the future package | Specifying more work than there are cores | How many cores should each process use? | Multiple cores per process | Other ways of using cores | Specifying which nodes should run your tasks | Memory requests | Running on specific nodes | Selecting by queue | Selecting by node names | Task time limits and scheduling | The maximum runtime | Delaying tasks starting | Lowering your priority

My task failed | Caused by an error in your code | Caused by an error during startup | My task got stuck at submitted | My code works on my computer but not on the cluster | Some of my tasks work on the cluster, but others fail | My code is slower on the cluster than running locally! | I can't connect to the cluster | Asking for help

Local installations | Latest available version of INLA for your R version | Specific version of INLA, where available. | Hipercow provisioning | Latest version of INLA for latest R version. | Latest version of INLA for older R version

Local installations | CmdStan on the cluster | Installation and versions | Compiling models | A worked example

Getting started | The lightweight queue pattern | Basic usage | Scaling up | Interprocess commuication pattern | Tricks and tips | Controlling the worker environment | General considerations | Stopping redundant workers | Permanence | Object storage

Impact by calendar year & impact by birth year | Function interface | Recipe interface | Impact by year of vaccination | Comparison | HepB | Measles | Comparisons of HepB, Measles & YF

Data | Impact by calendar year (cross-sectional impact) | Impact by birth year (lifetime impact) | Impact by year of vaccination | Impact by year of vaccination: impact ratio stratified by activity type | Impact by year of vaccination: impact ratio stratified by birth cohort | Glossary

Combining schemas

Motivation | Choropleth | Barchart | Table | Implementation overview | Filters | Plot Settings Control | Settings | Effects | stateFilterId vs filterId | Typescript definition and derived frontend store | Selections state | Data state | Endpoints | Front end implementation | Overview | Data fetching | Metadata fetching | On mount | On plot control/filter update | Adding new plot | Where do I write custom logic?

Where is everything? | The particle filter | Other state space methods | Running MCMC

Some simple examples | Calculations in the DSL | Pass in fixed data

Comparisons | The bendy banana | Sampling with other samplers

The structure of monty samples | Conversion to posterior's draw objects | Conversion to coda's mcmc.list objects

0. Prepare webtool GeoJSON input | 1. (Up)Load data inputs | 2. Choose model areas and time points | 3. Review input data | 4. Prepare model inputs

Dude, where's my feature? | Change in meaning of time in discrete-time models | New interface | See also

Error handling | Increasing resilience via separate processes | Loss of workers | Retrying tasks | An example where retrying fixes an error | Performance considerations | Deleting tasks that have been retried | Going further

Messages and responses | PING | ECHO | INFO | EVAL | PAUSE / RESUME | SET_TIMEOUT / GET_TIMEOUT | Messages that are supported but use via wrappers:

Getting started | Hello world | Design | Running tasks | Configuring the worker environment | Scheduling options | Tasks that depend on other tasks | Multiple queues | Running tasks in separate processes | Coping with memory use | Orchestrating workers | Use (and limitations) of spawn | Start workers on another node, perhaps using a scheduler | Use docker | Waiting for workers to appear | Worker heartbeat | Getting a Redis server | Install Redis | Use Redis on a different machine

Using print() | print format strings | Conditional display | Controlling precision | Current limitations | Show the generated code | Use the interactive debugger

A simple example - random walk | Running a model in parallel | A more interesting example | Other methods | Reordering particles | Set particle state

Influential environment variables and options

Running multiple realisations | Parallelisation | Efficient running | Efficient state handling | Useful verbs | A compilation target

The basic idea | An example

Introduction | Using set_antimalarial_resistance() to parameterise resistance | Simulating static resistance | Parameterisation | Simulation | Visualisation | Simulating dynamic resistance | Simulating antimalarial resistance with multiple resistance outcomes | References

Age group rendering | Set custom demography | Simulation | Visualisation | Set dynamic demography | Parameterisation

Mass Drug Administration | Parameterisation | Interventions | Simulation | Visualisation | Seasonal Malaria Chemoprevention | Simulations

PfPR~2-10~ matching using malariaEquilibrium | PfPR~2-10~ matching using malariasimulation | Establish malariasimulation parameters | Run the simulations and calculate PfPR~2-10~ | Fit line of best fit relating initial EIR values to PfPR~2-10~ | Visualisation | Matching EIR to PfPR~2-10~ values | Calibrating PfPR~2-10~ using the cali package

Run metapopulation simulation | Parameterisation | Simulation | Visualisation

Model structure | Human Biology | States | Parameters | Mosquito Biology | Key Model References (structure and dynamics) | Run simulation | Code | Output | Additional outputs | Output visualisation | Set equilibrium | Override parameters | Seasonality | Individual mosquitoes | Vignettes

Seasonality | Custom modification of the mosquito carrying capacity | Larval source management | Invasive species | Full flexibility

Setting mosquito species parameters | Single endophilic mosquito species | Single exophilic mosquito species | Plot adult female infectious mosquitoes by species over time | Setting multiple mosquito species

Setting drugs and clinical treatment | Parameterisation and simulation | Visualisation

Parameterisation | Mass RTS,S | Simulation | Visualisation | Seasonal mass vaccination | RTS,S EPI | RTS,S EPI with seasonal boosters | RTS,S dosing | TBV

Setting IRS parameters | Parameterisation | A note on mosquito species | Simulation | Visualisation

Discrete time systems | Continuous time systems | Properties of resettable variables | Unresolved bits that might change

Introduction | Bitset | Prefabs | C++ Prefabs

Running the model without vaccination | Running the model with vaccination | Graphical representations of results

Setup | Running the MCMC | Exploring outputs and checking MCMC performance | Using C++ functions

Model | Single temperature rung (no Metropolis coupling) | Multiple temperature rungs

A simple model of population growth

Installing Rcpp | Installing and loading drjacoby

Setup | Running the MCMC | How many rungs to use?

Setup | Running multiple chains | Running multiple chains using C++ log likelihood or log prior functions

Installation | Example 1: A simple population without specified population size | Population setup | Add vaccination activities | Visualisation | Example 2: A population with specified size and age distribution | Summary | Infer vaccination activities from popim_population object

Data structures | S3 class popim_population | The dataframe | Class attributes | Create, modify and validate objects of this class | S3 class popim_vacc_activities | Relationship between coverage, doses and target population | Create and validate objects of this class | Primary functionality: applying vaccination activities to a population | The inverse: inferring vaccination activities from a population | Population summary & visualisation | Visualisation | Aggregation over age

Setting seed and RNG type | Random variates with uniform distribution | Random variates with normal distribution | Random variates with exponential distribution | Random variates with Rademacher distribution | Random sampling | Getting and setting the RNG state | Accessing the global RNG

Threefry: usage from R | Xo(ro)shiro: jump ahead with OpenMP | PCG: multiple streams with RcppParallel | Using the global RNG

Supported Random Number Generators | Usage from R | Usage from C++ | Using the compiled library functions | Using the header only library | Accessing the global RNG

Introduction | Usage | Practical example | Caveats | Restoring random number generator state | Saving and restoring events

Handling lack of vault gracefully | Installing vault

Connecting to vault | Reading, writing, listing and deleting secrets | Alternative login approaches | Username and password (userpass) | GitHub (github) | LDAP (ldap)

Avoiding output() | Limited support for interpolate() | Delays are not supported | Not all distributions are supported | Interface differences

Using MCMC to infer parameters

Benchmarks | Technicalities

The packages | Documentation | The TypeScript ecosystem | Roadmap

Coping with missing data

Single variable: Logistic growth | Specifying parameters | More than one variable: the Lorenz attractor | Delay models | Arrays | Generalised Lotka-Volterra model | Interpolating functions

Ordinary differential equation models | Models implemented in R | Dense output and interpolation | Delay differential equation models | A model in R | Differences with deSolve | Models implemented in C

A simple example - random walk | Model code | Constructing a model | Running a model in parallel | A more interesting example | Other methods | Reordering particles | Set particle state | Reset the model | Use within a package

Background using R's random number generator | Basic implementation using dust | Parallel implementation with dust and OpenMP | More on the pointer object

Summary table of other R packages and tools

Running multiple realisations | Parallelisation | Efficient running | Efficient state handling | Useful verbs | A compilation target

Using print() | print format strings | Conditional display | Controlling precision | Current limitations

Setup | All-in-one | Restarting

Stochastic SIR model definition | Inferring parameters with mcstate | Model data | Defining the comparison function | Inferring parameters | Using MCMC to infer parameters | Tuning the pMCMC | Running predictions | Fitting to multiple datastreams

Stochastic SIR model definition | Implementing the SIR model using odin.dust | Saving a model into a package | Running the SIR model with dust | Adding age structure to the model

A note on seeding | Distributed seeding | Continuing the streams | Considerations | Use cases | Summary

Principles | Running a model with GPU support | Writing a GPU-capable model | Data comparison functions | Developing a GPU model

Supported distributions | Performance | Underlying random number engine | Reusing the random random number generator in other projects | In a package | Standalone, parallel with OpenMP | Standalone, parallel on a GPU | Other packages with similar functionality

Discrete compartmental models in a nutshell | From continuous to discrete time | Stochastic processes | Binomial distribution | Poisson distribution | Multinomial distribution | Implementation using odin | Deterministic SIR model | Stochastic SIR model | A stochastic SEIRDS model

Basic operators | Array support | Operators | Mathematical operators | Stochastic models

Introduction | Resizing variables | Birth and death processes | Simulation

General idea | Create demo package | Import function | Adding test with mock | Run individual tests | Write wrapper functions | Mock S3 methods

Problem motivation | Defining blocks | The Likelihood

Setting up an IF2 run | Running the IF2 algorithm

Model data | Comparison, model and particle filter | Nested parameters | pMCMC and Visualisations

Within-model parallelism | Between-chain parallelism | Considerations

Box-Muller | Polar | Ziggurat | Sampling | Sampling from the tail | The edges | Optimisations

Prerequisites | Declaring a simple endpoint | The basic syntax | Adding inputs | Binding state | Testing | An example

Frequently used functions communicating with Montagu database | Annex views (Susy to add) | Vaccine impact calculations

Considerations

Model | MCMC | Plots

Model | MCMC | Posterior plots

Model | Run MCMC | Plots

GET endpoint, inputs as query parameters, returning JSON | GET endpoint, inputs as path and query parameters, returning JSON | POST endpoint, inputs as JSON, returning JSON | POST endpoint, inputs as binary, returning binary

Motivation | Introduction - a porcelain approach to adding-as-a-service

Introduction | Variables | Categorical Variable | Integer Variable | Double Variable | Processes | Render | Events | Targeted Events | Simulate

Introduction | Specification | Processes | Events | Rendering | Simulation

Interpolation | Pluralisation | Fallback language | Translating multiple keys in a block of text | Use within a package | Namespaces, and the structure of translation files

Set up data | Incidence | Serial Interval Distribution

Data model diagramme | Areas data | Population data | Survey data | Programme data

Identify the Montagu Server | Basic Information | Touchstones | Scenarios | Expectations - years, ages, countries and outcomes | Demography | Gender | Format | Templates | Coverage | Long or Wide format | Returning the full range of countries | Burden Estimate Sets | Creating a new burden estimate set | Uploading data | Plotting previous burden estimate sets | Stochastic Runs | Model Run Parameter Sets | Appendix | Models | Demographic Sources

Package preparation | The API

Files accessed by specio | .DP | .xml | .SPT - EPP results for Spectrum | Section 1: National results | Section 2: National F/M ratios and IDU info | Section 3: Total workset results | Section 4: Workset F/M ratios and IDU info | Section 5: First sub-population results | Section 6: First sub-population F/M ratios and IDU info | General notes | .SPU - EPP uncertainty results for Spectrum | Files not currently accessed by specio | ep1 - year and demographic inputs that EPP needs from Spectrum | typ - type of the EPP workset either GENERALIZED or CONCENTRATED | ep4 - ART data and parameters

Introduction | Background | Usage - how? | Usage - why?