Prediction and simulation method for 'mobility.model' class

This function uses a fitted mobility.model object to simulate a connectivity matrix based on estimated parameters.

predict(object, newdata, nsim, seed, ...)

Arguments

object	a mobility.model object containing a fitted mobility model and its associated data
newdata	a list containing new data to used in model prediction. If NULL (the default) the function will simulate the model with the data used for fitting. D a named matrix of distances between all \(ij\) location pairs N a named vector of population sizes for all locations (either N or both N_orig and N_dest must be supplied) N_orig a named vector of population sizes for each origin N_dest a named vector of population sizes for each destination
nsim	number of simulations (default = 1).
seed	optional integer specifying the call to set.seed prior to model simulation (default = NULL)
...	further arguments passed to or from other methods

Value

a vector, matrix, or array containing predited or simulated mobility values.

Details

When nsim = 1, the prediction matrix is calculated using the mean point estimate of parameter values. If nsim > 1 then returns and array that contains nsim number of simulated replications based on the posterior distributions of each parameter.

See also

Other model: check(), compare(), fit_jags(), fit_prob_travel(), mobility(), residuals(), summary()

Author

John Giles

Examples

mod <- mobility(data=mobility_matrices,
                model='gravity',
                type='transport')
#> ::Fitting transport gravity model for 10 origins and 10 destinations::
#> Compiling model graph
#>    Resolving undeclared variables
#>    Allocating nodes
#> Graph information:
#>    Observed stochastic nodes: 70
#>    Unobserved stochastic nodes: 32
#>    Total graph size: 417
#> 
#> Initializing model
#> 
#> NOTE: Stopping adaptation
#> 
#> 

mod <- mobility(data=mobility_matrices,
                model='departure-diffusion',
                type='power',
                hierarchical=TRUE)
#> ::Fitting power departure-diffusion model for 10 origins and 10 destinations::
#> Compiling model graph
#>    Resolving undeclared variables
#>    Allocating nodes
#> Graph information:
#>    Observed stochastic nodes: 70
#>    Unobserved stochastic nodes: 46
#>    Total graph size: 994
#> 
#> Initializing model
#> 
#> NOTE: Stopping adaptation
#> 
#> 

predict(object=mod)
#>       destination
#> origin          A          B           C           D          E          F
#>      A 4612.33342  348.37077  125.368853  509.947769  336.41805  238.15704
#>      B   31.78730 8974.50837   51.627433   13.305170   71.53423  132.24174
#>      C   13.52665   61.04770 5283.612608    7.233958   49.45510   68.34040
#>      D 1926.71267  550.93429  253.318307 2270.236594  636.69850  417.66652
#>      E  267.36113  623.04722  364.275050  133.924965 8604.64752 1273.56705
#>      F   44.56248  271.18309  118.517585   20.684467  299.85325 6408.55208
#>      G  200.69147  118.50892   31.293662   41.993366   89.10860   71.80827
#>      H  150.10473  113.30474   27.701293   34.133658   67.80793   59.57853
#>      I   33.00185   22.61175    5.647168    7.216220   14.22255   12.26176
#>      J  731.95688  434.33209  215.208769  507.052971  862.45931  415.97965
#>       destination
#> origin          G          H          I           J
#>      A 1320.30306 1307.65149  785.70147  272.417172
#>      B   71.13898   90.06532   49.12078   14.749687
#>      C   22.21272   26.03744   14.50610    8.641907
#>      D 1043.79846 1123.49652  649.11292  713.006632
#>      E  465.88992  469.45837  269.10124  255.097714
#>      F   88.39448   97.11658   54.62331   28.968566
#>      G 7283.78274 1262.68985  977.93772   41.552495
#>      H  953.55064 4125.38938 6495.92529   29.643308
#>      I  270.23193 2376.94654 2849.07131    6.354144
#>      J  734.50175  693.86556  406.46893 1584.602553
#> attr(,"model")
#> [1] "departure-diffusion"
#> attr(,"type")
#> [1] "power"
#> attr(,"hierarchical")
#> [1] TRUE

n <- 5
ids <- letters[1:n]

# Distance matrix
D <- get_distance_matrix(x=rnorm(n, -100, 2),
                         y=rnorm(n, 20, 1),
                         id=ids)*111.35

# Vector of population sizes
N <- rnbinom(n, size=5, mu=5000)
names(N) <- ids

# Predict mobility model using new data
predict(object=mod, newdata=list(D=D, N=N))
#>       destination
#> origin          a          b          c         d          e
#>      a 3459.71747  626.90386 1712.31675 390.48386 1204.07873
#>      b   33.88195 2872.29560   45.18656  71.13985   18.00077
#>      c   74.68437   36.46594 5419.47965  25.02005  141.80039
#>      d  342.24054 1153.64941  502.77192 687.26941  215.73305
#>      e  247.35776   68.42177  667.88622  50.56604 2159.09725
#> attr(,"model")
#> [1] "departure-diffusion"
#> attr(,"type")
#> [1] "power"
#> attr(,"hierarchical")
#> [1] TRUE