NEWS.md

0.10.7

• non-standard parameters from custom Bayesian models are now sorted in a less bewildering order in the model fit (as from get_fit())

0.10.4

• plot_DO_preds with dygraphs works again, and now you can subset dates using args date_start and date_end directly in plot_DO_preds

0.10.3

• adding a tiny bit of support toward bringing in new light-saturation models in the future

0.10.2

• test for day lengths < 0 in mm_model_by_ply, thanks to @weisoon

• bugfix for get_params for sim models with fixed vector of Q and/or K values

0.10.1

• upgrade unitted dependency to v0.2.8 to accommodate recent change in function exporting requirements in R 3.3

• increase robustness of mm_is_valid_day for dates with 1 observation

0.10.0

• more comprehensive assignment of parameters from Stan output into model fit data.frames in Bayesian models

• daily and inst data.frames in Bayesian model fit now get date/timestamps

• structural changes in Bayesian models: fitted DO_mod[1] for state space models, probability constraints on DO_mod[1] and DO_mod_partial[1]

• structural change in Bayesian models: reindexing err_obs_iid and err_proc_iid to match other inst variables

0.9.48

• Extend filtering with required_timestep to metab_bayes

0.9.47

• Bug fix in use of required_timestep in mm_is_valid_day

0.9.46

• Removed the warning discouraging setting params_out within specs()

0.9.45

• Bayesian specs defaults now reflect a little more of the literature and our experience modeling metabolism

0.9.44

• New specs element required_timestep allows you to require that each date has the specified numeric timestep in days

0.9.42-0.9.43

• Bugfixes for new test that excludes days with non-positive discharge

0.9.40

• Update to roxygen2 6.0.1

0.9.39

• Bayesian, MLE, and nighttime regression models can now all test for and exclude days with non-positive discharge

0.9.36

• Models should now be able to accept tbl_dfs (dplyr/tibble format) for the data and data_daily arguments to metab()

0.9.35

• Bayesian models now distinguish between compilation time and fitting time

• Updates to plot_distribs for recent changes to Bayesian models

0.9.34

• Bayesian models with pool_K600 != 'none' can how have their K600_daily_sigma (or K600_daily_sdlog) be a fitted value, a fixed value, or a value fixed at 0

0.9.33

• new function: calc_light_merged, which merges modeled and observed light into a smooth curve

0.9.29

• functions in the specs of sim models can now refer to their own metabolism model (and therefore also its info or data_daily slots)

0.9.28

• switched from rlnorm to rnorm for distribution of K600_daily around K600_daily_pred in linear and binned models

0.9.27

• sim models can now generate daily parameters from functions and can even generate binned K~Q relationships with random variation

0.9.25

• various improvements to flexibility and speed of sim models

0.9.22

• renamed metab_sim specs for consistency with other model arguments, e.g., err.obs.sigma is now err_obs_sigma

0.9.21

• simplified Bayesian K pooling models to make the pooling more effective (and also faster). Hierarchical bayesian models now fix rather than fit K600_daily_sdlog, the standard deviation of K600_daily relative to K600_daily_predlog

0.9.18

• solidified the model feature options, including error types, DO deficit source, ODE integration methods, and hierarchy

• revised hierarchical methods to be consistent with recent email exchanges.

0.9.15

• new function: plot_distribs to explore priors

• bayesian models can now track err_obs_iid and other parameters fit at the resolution of data

• new functions: calc_solar_time and calc_light for faster data preparation

• refined treatment of depth in trapezoid method - effect will mainly be noticeable for sites/times with rapidly changing depth

0.9.14

• all tests are passing and all examples are running [again]. tests are acceptably comprehensive [for now].

• known issue: K binning in bayesian models is producing unlikely estimates. all bayesian models need further testing.

0.9.13

• ongoing bug fixes and re-integration of all model types into new ODE production/integration scheme

0.9.12

• removed JAGS from the package. Stan is better for us.

0.9.11

• this is the last version with JAGS in it.

0.9.10

• myriad bug fixes to accommodate changes from version 0.9.9.

0.9.9

• metab_mle and metab_sim now have several options for relationships among GPP, light, ER, and temperature.

• metab_mle and metab_sim now have several options for the method by which the differential equation for dDO/dt is numerically integrated to produce a time series of DO predictions.

• metab_mle has taken a performance hit to become more flexible in the GPP-light relationship, the ODE method, and so on. Optimization is likely in the future.

• predict_metab and predict_DO optionally attach units to their output.

0.9.8

• better error and warning handling in metab_bayes, including a new function get_log() that retrieves log file[s] from MCMC model compilation & run[s]. log files are now retrieved for both JAGS and Stan models.

• updates to accommodate changes in dependency packages (tibble and dplyr)

• more efficient specifications of JAGS and Stan models

• incorporated feedback on vignette

0.9.6

• model names (from mm_name()) and bayesian model file names (in models folder) now include info on the GPP and ER functions - default is still pl = GPP is a linear function of light, and rc = ER is constant over every 24-hour period

• more informative error messages for timesteps <= 0 in mm_model_by_ply

0.9.5.1

• in metab_Kmodel, now avoiding negative weights

0.9.5

• Bug fixes and error prevention

0.9.4

• Now automatically checks for available updates when you attach the package

• Improved units handling in convert_k600/kGAS

0.9.2

• Hierarchical constraints on K600 are now available! Options are 'normal', 'linear', and 'binned'; see the details section on pool_K600 in ?mm_name and the description of parameters starting with K600_daily in ?specs.

• Interface change: specs lists now print more prettily and have class 'specs' (though they're still fundamentally just lists)

• Vignette: see vignette('getstarted')

0.9.0

• New function: metab() serves as a gateway to all model types. You can now pass specs to metab() and expect the appropriate model to be chosen and called based on model_name in the specs list.

• New function: data_metab() produces a dataset for testing/demonstration, with options for the resolution & flaws to introduce.

• Newly public function: mm_model_by_ply is now public. Its interface has also changed somewhat: tests has been renamed to day_tests, and validity tests are conducted within mm_model_by_ply if day_tests is not empty, and validity and timestep information are now passed to model_fun.

• Changed functionality: mm_model_by_ply_prototype() now produces a 1-row data.frame as well as a message, which helps this function demonstrate the workings of mm_model_by_ply(). mm_model_by_ply_prototype() is a lightweight example of a function that can be passed to mm_model_by_ply(), and its help file describes the parameters such a function should expect.

• New function: mm_get_timestep() computes the mean and/or unique timestep[s] and optionally requires that there be just one unique timestep within a vector of times or dates.

• Interface change: the argument tests is now called day_tests in the metab(), metab_night(), etc., mm_model_by_ply(), and mm_is_valid_day().

• Interface change: day_start, day_end, and tests are now containined within specs rather than defined separately in the call to metab, metab_bayes, etc.

• Interface change: in metab(), metab_mle(), etc., the model_specs argument is now called specs.

• Interface change: metab functions now accept specs first, then data, data_daily, and info. (specs was renamed from model_specs; see above.) This permits chaining from mm_name to specs to metab.

• Interface change: get_args is now get_specs, and the result is a list of specs as named in specs() rather than a list with an element called model_specs that is itself a list.

• Hierarchical bayesian models are now possible and include hierarchical parameters for distributions on error and K600 (normal, linear, and binned). Some models are known to work; complete testing for all models is forthcoming.

0.8.0

• Major interface change (renamed variable) to clarify types of time: solar.time (mean solar time), app.solar.time (apparent solar time), local.time (time in local time zone). Metabolism models now accept solar.time rather than local.time, though it's still possible to pass in local time but just call it solar.time (as long as you don't have daylight savings time).

0.7.3

• Remove calc_schmidt because it is never used

0.7.2

This package is not ready for use by many, but it does currently have:

• support for a wide range of non-hierarchical models, both Bayesian and MLE-based

• support for regressions of daily K versus discharge and/or velocity

• default specifications for every model

• a maturing user interface for fitting models (probably not quite fixed yet)

• convenience functions for calculating DO saturation concentrations, air pressure, depth, solar time, PAR, etc.

• functions for simulating data and error, for testing models with data having known underlying parameters

• two small datasets, courtesy of Bob Hall, for testing models with real data