test_that("model-assessment.R works on with GEV/L-moments/S-bootstrap.", {

	# Use identical parameters and uncertainty results to the MATLAB version 
	ci <- data.frame(
		periods = c(2, 5, 10, 20, 50, 100),
		lower = c(1678.8540, 2361.9491, 2821.1799, 3246.2090, 3764.1510, 4134.4393),
		upper = c(1997.5810, 2883.5746, 3602.7005, 4449.9595, 5840.9811, 7144.6336) 
	)

	# Load dataset and run L-moments estimation with uncertainty analysis
	df <- data_local("CAN-07BE001.csv")
	metrics <- model_assessment(df$max, "GEV", "L-moments", ci = ci)$metrics

	# Test results against MATLAB
	expect_equal(metrics$R2      ,    0.9922, tol = 1e-4)
	expect_equal(metrics$RMSE    ,   95.9225, tol = 1e-4)
	expect_equal(metrics$bias    ,  -20.6072, tol = 1e-4)
	expect_equal(metrics$AIC_RMSE,  471.4811, tol = 1e-4)
	expect_equal(metrics$BIC_RMSE,  479.3561, tol = 1e-4)
	expect_equal(metrics$AIC_MLL , 1656.8821, tol = 1e-4)
	expect_equal(metrics$BIC_MLL , 1664.7570, tol = 1e-4)
	expect_equal(metrics$AW      ,  603.0077, tol = 1e-4)
	expect_equal(metrics$POC     ,  100.0000, tol = 1e-4)
	expect_equal(metrics$CWI     ,  545.6239, tol = 1e-4)

})

test_that("model-assessment.R works with WEI/L-moments/S-bootstrap.", {

	# Use identical parameters and uncertainty results to the MATLAB version 
	ci <- data.frame(
		periods = c(2, 5, 10, 20, 50, 100),
		lower = c(1430.3752, 2058.0287, 2391.4038, 2656.8127, 2944.4725, 3130.8841),
		upper = c(1759.9254, 2462.2588, 2895.8749, 3295.6570, 3790.4480, 4141.0544) 
	)

	# Load dataset and run L-moments estimation with uncertainty analysis
	df <- data_local("CAN-08NH021.csv")
	metrics <- model_assessment(df$max, "WEI", "L-moments", ci = ci)$metrics

	# Test results against MATLAB
	expect_equal(metrics$R2      ,    0.9862, tol = 1e-4)
	expect_equal(metrics$RMSE    ,   81.8381, tol = 1e-4)
	expect_equal(metrics$bias    ,   -6.6950, tol = 1e-4)
	expect_equal(metrics$AIC_RMSE,  406.8316, tol = 1e-4)
	expect_equal(metrics$BIC_RMSE,  414.3641, tol = 1e-4)
	expect_equal(metrics$AIC_MLL , 1442.7845, tol = 1e-4)
	expect_equal(metrics$BIC_MLL , 1450.3171, tol = 1e-4)
	expect_equal(metrics$AW      ,  440.2948, tol = 1e-4)
	expect_equal(metrics$POC     ,   97.7778, tol = 1e-4)
	expect_equal(metrics$CWI     ,  416.5010, tol = 1e-4)

})

test_that("model-assessment.R works with GNO/L-moments/S-bootstrap.", {

	# Use identical parameters and uncertainty results to the MATLAB version 
	ci <- data.frame(
		periods = c(2, 5, 10, 20, 50, 100),
		lower = c(64.2910, 82.0855,  92.8011, 102.0788, 113.0530, 120.7327),
		upper = c(73.4937, 95.4063, 111.4363, 128.3836, 152.6660, 172.5685) 
	)

	# Load dataset and run L-moments estimation with uncertainty analysis
	df <- data_local("CAN-08MH016.csv")
	metrics <- model_assessment(df$max, "GNO", "L-moments", ci = ci)$metrics

	# Test results against MATLAB
	expect_equal(metrics$R2      ,   0.9884, tol = 1e-4)
	expect_equal(metrics$RMSE    ,   2.6046, tol = 1e-4)
	expect_equal(metrics$bias    ,  -0.3580, tol = 1e-4)
	expect_equal(metrics$AIC_RMSE,  91.1979, tol = 1e-4)
	expect_equal(metrics$BIC_RMSE,  98.6638, tol = 1e-4)
	expect_equal(metrics$AIC_MLL , 791.5803, tol = 1e-4)
	expect_equal(metrics$BIC_MLL , 799.0462, tol = 1e-4)
	expect_equal(metrics$AW      ,  15.4410, tol = 1e-4)
	expect_equal(metrics$POC     , 100.0000, tol = 1e-4)
	expect_equal(metrics$CWI     ,  13.9716, tol = 1e-4)

})

test_that("model-assessment.R works with LP3/L-moments/S-bootstrap.", {

	# Use identical parameters and uncertainty results to the MATLAB version 
	ci <- data.frame(
		periods = c(2, 5, 10, 20, 50, 100),
		lower = c(31.2190, 47.9695, 57.7789, 65.6528, 73.6143, 78.2450),
		upper = c(39.5256, 59.6762, 73.0359, 86.4724, 105.1122, 120.1799) 
	)

	# Load dataset and run L-moments estimation with uncertainty analysis
	df <- data_local("CAN-08NM050.csv")
	metrics <- model_assessment(df$max, "LP3", "L-moments", ci = ci)$metrics 

	# Test results against MATLAB. 
	expect_equal(metrics$R2      ,   0.9669, tol = 1e-2)
	expect_equal(metrics$RMSE    ,   3.4106, tol = 1e-2)
	expect_equal(metrics$bias    ,  -0.0061, tol = 1e-2)
	expect_equal(metrics$AIC_RMSE, 125.0092, tol = 1e-2)
	expect_equal(metrics$BIC_RMSE, 132.7334, tol = 1e-2)
	expect_equal(metrics$AIC_MLL , 833.1819, tol = 1e-4)
	expect_equal(metrics$BIC_MLL , 840.9061, tol = 1e-4)
	expect_equal(metrics$AW      ,  13.1347, tol = 1e-4)
	expect_equal(metrics$POC     ,  85.4167, tol = 1e-4)
	expect_equal(metrics$CWI     ,  15.9096, tol = 1e-4)

})

test_that("model-assessment.R works with nonstationary GEV/MLE/S-bootstrap.", {

	set.seed(1)
	df <- data_local("CAN-07BE001.csv")
	metrics <- model_assessment(df$max, "GEV", "MLE", NULL, df$year, S10)$metrics

	# Test results against MATLAB 
	expect_equal(metrics$R2      , 0.9865   , tol = 1e-4)
	expect_equal(metrics$RMSE    , 0.1210   , tol = 1e-4)
	expect_equal(metrics$bias    , 0.0008   , tol = 1e-4)
	expect_equal(metrics$AIC_RMSE, -207.384 , tol = 1e-4)
	expect_equal(metrics$BIC_RMSE, -196.884 , tol = 1e-4)
	expect_equal(metrics$AIC_MLL , 1657.8901, tol = 1e-4)
	expect_equal(metrics$BIC_MLL , 1668.3900, tol = 1e-4)
	expect_equal(metrics$AW      , NULL)
	expect_equal(metrics$POC     , NULL)
	expect_equal(metrics$CWI     , NULL)

})