Skip to content

Commit

Permalink
update readme and vignettes
Browse files Browse the repository at this point in the history
  • Loading branch information
@nicholasjclark
nicholasjclark committed Sep 24, 2023
1 parent f115d4e commit 061dded
Show file tree
Hide file tree
Showing 92 changed files with 327 additions and 320 deletions.
6 changes: 5 additions & 1 deletion R/update.mvgam.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -93,7 +93,11 @@ update.mvgam = function(object, formula,
}

if(missing(priors)){
priors <- object$priors
if(!is.null(object$priors)){
priors <- object$priors
} else {
priors <- rlang::missing_arg()
}
}

if(!missing(newdata)){
Expand Down
76 changes: 38 additions & 38 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -347,11 +347,11 @@ test_priors
#> 6 trend sd sigma ~ student_t(3, 0, 2.5);
#> example_change new_lowerbound new_upperbound
#> 1 (Intercept) ~ normal(0, 1); NA NA
#> 2 lambda ~ exponential(0.82); NA NA
#> 3 ar1 ~ normal(-0.99, 0.33); NA NA
#> 4 ar2 ~ normal(0.18, 0.76); NA NA
#> 5 ar3 ~ normal(0.48, 0.15); NA NA
#> 6 sigma ~ exponential(0.52); NA NA
#> 2 lambda ~ exponential(0.08); NA NA
#> 3 ar1 ~ normal(0.71, 0.3); NA NA
#> 4 ar2 ~ normal(0.18, 0.25); NA NA
#> 5 ar3 ~ normal(0.75, 0.68); NA NA
#> 6 sigma ~ exponential(0.51); NA NA
```

Any of the above priors can be changed by modifying the `prior` column
Expand Down Expand Up @@ -552,33 +552,33 @@ summary(lynx_mvgam)
#> Fitted using Stan
#>
#> GAM coefficient (beta) estimates:
#> 2.5% 50% 97.5% Rhat n.eff
#> (Intercept) 5.900 6.600 7.000 1.00 323
#> s(season).1 -0.630 0.030 0.740 1.00 1080
#> s(season).2 -0.240 0.820 1.800 1.00 467
#> s(season).3 -0.055 1.200 2.500 1.00 355
#> s(season).4 -0.590 0.430 1.400 1.00 1074
#> s(season).5 -1.200 -0.160 0.910 1.00 506
#> s(season).6 -1.100 -0.031 1.000 1.00 523
#> s(season).7 -0.690 0.360 1.400 1.00 769
#> s(season).8 -0.990 0.290 1.800 1.00 340
#> s(season).9 -1.100 -0.240 0.720 1.01 392
#> s(season).10 -1.300 -0.660 -0.025 1.00 764
#> 2.5% 50% 97.5% Rhat n.eff
#> (Intercept) 6.000 6.600 7.00000 1.01 408
#> s(season).1 -0.630 0.033 0.76000 1.01 823
#> s(season).2 -0.230 0.850 1.90000 1.01 382
#> s(season).3 -0.065 1.200 2.50000 1.01 358
#> s(season).4 -0.530 0.430 1.40000 1.01 856
#> s(season).5 -1.200 -0.160 0.91000 1.00 551
#> s(season).6 -1.100 -0.012 1.00000 1.00 625
#> s(season).7 -0.770 0.360 1.40000 1.00 614
#> s(season).8 -1.100 0.270 1.80000 1.00 374
#> s(season).9 -1.100 -0.310 0.68000 1.00 410
#> s(season).10 -1.400 -0.700 0.00064 1.01 574
#>
#> GAM observation smoothing parameter (rho) estimates:
#> 2.5% 50% 97.5% Rhat n.eff
#> s(season)_rho 2.1 3.4 4.3 1 415
#> s(season)_rho 2.1 3.4 4.2 1 520
#>
#> Approximate significance of GAM observation smooths:
#> edf Ref.df Chi.sq p-value
#> s(season) 6.56 10 19358 0.24
#> edf Chi.sq p-value
#> s(season) 4.04 20290 0.22
#>
#> Latent trend AR parameter estimates:
#> 2.5% 50% 97.5% Rhat n.eff
#> ar1[1] 0.74 1.10 1.400 1 614
#> ar2[1] -0.82 -0.39 0.065 1 1524
#> ar3[1] -0.46 -0.12 0.320 1 432
#> sigma[1] 0.40 0.50 0.650 1 1216
#> ar1[1] 0.76 1.10 1.400 1.00 747
#> ar2[1] -0.85 -0.40 0.062 1.00 1608
#> ar3[1] -0.50 -0.12 0.330 1.01 470
#> sigma[1] 0.40 0.50 0.640 1.00 1031
#>
#> Stan MCMC diagnostics:
#> n_eff / iter looks reasonable for all parameters
Expand Down Expand Up @@ -688,7 +688,7 @@ plot(lynx_mvgam, type = 'forecast', newdata = lynx_test)
<img src="man/figures/README-unnamed-chunk-31-1.png" width="60%" style="display: block; margin: auto;" />

#> Out of sample DRPS:
#> [1] 2806.193
#> [1] 2896.107

And the estimated latent trend component, again using the more flexible
`plot_mvgam_...()` option to show first derivatives of the estimated
Expand Down Expand Up @@ -788,9 +788,9 @@ each competing model at the same set of timepoints.
``` r
compare_mvgams(lynx_mvgam, lynx_mvgam_poor, fc_horizon = 10)
#> RPS summaries per model (lower is better)
#> Min. 1st Qu. Median Mean 3rd Qu. Max.
#> Model 1 3741.261 4251.22 4740.237 5170.38 5900.935 7639.796
#> Model 2 5625.613 11398.04 16985.819 15268.62 19268.205 21496.856
#> Min. 1st Qu. Median Mean 3rd Qu. Max.
#> Model 1 3726.048 4241.425 4879.843 5248.287 5995.075 7779.446
#> Model 2 5644.477 11567.963 17900.513 15766.338 19783.975 22561.216
#>
#> 90% interval coverages per model (closer to 0.9 is better)
#> Model 1 0.98
Expand Down Expand Up @@ -904,9 +904,9 @@ Total ELPDs per model are also a useful overall indicator of performance

``` r
lfo_good$sum_ELPD
#> [1] -134.3953
#> [1] -141.4854
lfo_poor$sum_ELPD
#> [1] -156.5703
#> [1] -166.5148
```

As before, these metrics all favour the more complex model over the
Expand Down Expand Up @@ -1011,11 +1011,11 @@ summary(mod, include_betas = FALSE)
#> s(season):seriesseries_3_rho 1.2 3.0 4.1 1 992
#>
#> Approximate significance of GAM observation smooths:
#> edf Ref.df Chi.sq p-value
#> s(season) 4.66 5 15.46 1.9e-05 ***
#> s(season):seriesseries_1 1.41 4 0.12 0.98
#> s(season):seriesseries_2 1.78 4 0.10 0.98
#> s(season):seriesseries_3 3.31 4 0.82 0.46
#> edf Chi.sq p-value
#> s(season) 5.00 15.46 1.9e-05 ***
#> s(season):seriesseries_1 3.74 0.12 0.98
#> s(season):seriesseries_2 3.83 0.10 0.98
#> s(season):seriesseries_3 3.97 0.82 0.46
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
Expand Down Expand Up @@ -1153,8 +1153,8 @@ summary(mod, include_betas = FALSE)
#> s(time):temp2_rho -1.4 1.5 3.8 1 1517
#>
#> Approximate significance of GAM observation smooths:
#> edf Ref.df F p-value
#> s(time):temp 24.4 25.9 86.3 <2e-16 ***
#> edf F p-value
#> s(time):temp 14.9 80.2 <2e-16 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
Expand Down Expand Up @@ -1198,7 +1198,7 @@ plot(mod, type = 'forecast', newdata = data_test)
<img src="man/figures/README-unnamed-chunk-58-1.png" width="60%" style="display: block; margin: auto;" />

#> Out of sample CRPS:
#> [1] 1.295975
#> [1] 1.281124

There are many more extended uses for `mvgam` models, including the
ability to fit dynamic factor processes for analysing and forecasting
Expand Down
Loading

0 comments on commit 061dded

Please sign in to comment.