pit_sweq: Compute PIT (probability integral transform) for sweq
In robertsy/assimilr: Local Ensemble Kalman Particle Filters

Compute PIT (probability integral transform) for sweq

pit_sweq(sweq_run, model_run, fig = FALSE,
  train = sweq_run$params["thres.rain"], loc = seq(1, sweq_run$ndim, by =
  10), times = sweq_run$freq * seq(0, sweq_run$duration/sweq_run$freq, by =
  3))

`sweq_run`	is the true model run returned by sweq_simulate
`model_run`	is the assimilation run returned by da_cycle
`train`	the threshold of rain observations
`loc`	the locations at which to compute the PIT (better not too close to avoid correlations)
`times`	at which to compute the PIT (like for loc, thin out to avoid correlated values)
`fig=T/F`	to plot the results or not

set.seed(1)
sweq_run <- sweq_simulate(5*60, 60, 300)
k <- 20
ens0 <- sweq_ens0(k, sweq_run, klag=1000)
l <- 5
block_run <- da_cycle(ens0, sweq_run, block_LEnKPF, l=l, block_size=l/2, 
                      taper=sweq_GC(sweq_run$ndim, l, 0.9), ndim=sweq_run$ndim,
                      get_partition = sweq_partition,
                      manifold_proj = sweq_proj)
pit_df <- pit_sweq(sweq_run, block_run, fig=TRUE, train=train)
pit_df %>% pit_sweq_plot()