Prediction not NA with disjoint/segmented pred
SoftwareLiteracy opened this issue · 1 comments
SoftwareLiteracy commented
As pointed out by @nonlinearnature, the current version (1.8) does not output NA in prediction rows in the presence of disjoint or segmented pred. The code makes predictions in these "gaps" using the available library vectors.
An example of the current behavior:
NOTE: This was run prior to rEDM 1.15. In 1.15 the legacy wrapper block_lnlp was deprecated.
> block <- data.frame( time=1:10, x=sin((1:10)/pi), y=cos((1:10)/pi) )
> out <- block_lnlp(block,lib=rbind(c(1,5),c(6,10)),tp=2,columns=c("x","y"),target_column = "x",stats_only = FALSE)
> out$model_output
Index Observations Predictions Pred_Variance Const_Predictions
1 1 0.31296 NaN NaN NaN
2 2 0.59448 NaN NaN NaN
3 3 0.81627 0.95936 0.003832 0.31296
4 4 0.95606 0.89728 0.011778 0.59448
5 5 0.99978 0.88276 0.014781 0.81627
6 6 0.94307 0.89594 0.031104 0.95606
7 7 0.79160 0.59326 0.058161 0.99978
8 8 0.56060 0.26670 0.076164 0.94307
9 9 0.27328 0.28246 0.104256 0.79160
10 10 -0.04149 0.36578 0.024815 0.56060
11 11 NaN 0.09529 0.041300 0.27328
12 12 NaN 0.17593 0.055783 -0.04149
The behavior of rEDM 0.7.4 with disjoint pred:
> out74 <- block_lnlp( block, lib=rbind(c(1,5),c(6,10)), tp=2, columns=c("x","y"), target_column = "x", stats_only = FALSE )
> out74 $ model_output[[1]]
time obs pred pred_var
1 3 0.81627 0.9594 0.003832
2 4 0.95606 0.8973 0.011778
3 5 0.99978 0.8828 0.014781
4 6 0.94307 NaN NaN
5 7 0.79160 NaN NaN
6 8 0.56060 0.2667 0.076164
7 9 0.27328 0.2825 0.104256
8 10 -0.04149 0.3658 0.024815
9 NaN NaN NaN NaN
10 NaN NaN NaN NaN
SoftwareLiteracy commented
To clarify the issue, it is not with disjoint lib, rather disjoint pred. Disjoint lib yields results equivalent to rEDM 0.7.4.
In rEDM 0.7.4, if lib is specified and pred is not, then pred is set equal to lib. This is why the example above has NaN in prediction rows 4 and 5. In the 1.x code disjoint pred is not supported. It can be specified, but the output does not remove prediction rows in a prediction gap.
To clarify and provide examples:
library( rEDM )
library( rEDM74 )
df = data.frame( time = 1:10, x = sin((1:10)/pi), y = cos((1:10)/pi) )
# block_lnlp : Embedded = TRUE
B = rEDM74::block_lnlp( df, lib = c(1, 5, 6, 10), pred = c(1, 10),
method = "simplex", tp = 2, columns = c("x","y"),
target_column = "x", stats_only = FALSE,
first_column_time = TRUE )
B $ model_output[[1]]
time obs pred pred_var
1 3 0.81627 0.9594 0.003832
2 4 0.95606 0.8973 0.011778
3 5 0.99978 0.8828 0.014781
4 6 0.94307 0.8959 0.031104
5 7 0.79160 0.5933 0.058161
6 8 0.56060 0.2667 0.076164
7 9 0.27328 0.2825 0.104256
8 10 -0.04149 0.3658 0.024815
9 NaN NaN NaN NaN
10 NaN NaN NaN NaN
rEDM::Simplex( dataFrame = df, lib = "1 5 6 10", pred = "1 10", Tp = 2,
columns = "x y", target = "x", embedded = TRUE )
time Observations Predictions Pred_Variance
1 1 0.31296 NaN NaN
2 2 0.59448 NaN NaN
3 3 0.81627 0.95936 0.003832
4 4 0.95606 0.89728 0.011778
5 5 0.99978 0.88276 0.014781
6 6 0.94307 0.89594 0.031104
7 7 0.79160 0.59326 0.058161
8 8 0.56060 0.26670 0.076164
9 9 0.27328 0.28246 0.104256
10 10 -0.04149 0.36578 0.024815
11 11 NaN 0.09529 0.041300
12 12 NaN 0.17593 0.055783
# simplex : embedded = FALSE
S = rEDM74::simplex( df[ , c('time','x') ],
lib = c(1, 5, 6, 10), pred = c(1, 10),
E = 2, tp = 2, stats_only = FALSE )
S $ model_output[[1]]
time obs pred pred_var
1 3 0.81627 NaN NaN
2 4 0.95606 0.5076 0.22429
3 5 0.99978 0.3787 0.17233
4 6 0.94307 0.4721 0.16575
5 7 0.79160 0.3880 0.12633
6 8 0.56060 0.3348 0.08463
7 9 0.27328 0.4779 0.26568
8 10 -0.04149 0.6781 0.12408
9 NaN NaN NaN NaN
10 NaN NaN NaN NaN
rEDM::Simplex( dataFrame = df, lib = "1 5 6 10", pred = "1 10",
Tp = 2, E = 2, columns = "x", target = "x", embedded = FALSE )
time Observations Predictions Pred_Variance
1 2 0.59448 NaN NaN
2 3 0.81627 NaN NaN
3 4 0.95606 0.5076 0.22429
4 5 0.99978 0.3787 0.17233
5 6 0.94307 0.4721 0.16575
6 7 0.79160 0.3880 0.12633
7 8 0.56060 0.3348 0.08463
8 9 0.27328 0.4779 0.26568
9 10 -0.04149 0.6781 0.12408
10 11 NaN 0.5719 0.24683
11 12 NaN 0.6466 0.22507
# Disjoint pred : not specifying pred sets pred to lib in rEDM 0.7.4
B2 = rEDM74::block_lnlp( df, lib = c(1,5,6,10), method = "simplex",
tp = 2, columns = c("x","y"),target_column = "x",
stats_only = FALSE,first_column_time = TRUE )
B2 $ model_output[[1]]
time obs pred pred_var
1 3 0.81627 0.9594 0.003832
2 4 0.95606 0.8973 0.011778
3 5 0.99978 0.8828 0.014781
4 6 0.94307 NaN NaN
5 7 0.79160 NaN NaN
6 8 0.56060 0.2667 0.076164
7 9 0.27328 0.2825 0.104256
8 10 -0.04149 0.3658 0.024815
9 NaN NaN NaN NaN
10 NaN NaN NaN NaN
# 1.x code does not segment based on pred
# Since results are not affected, one can simply partition results
S2 = rEDM::Simplex( dataFrame = df, lib = "1 5 6 10", pred = "1 5 6 10",
Tp = 2,columns = "x y", target = "x", embedded = TRUE )
S2[ 6:7, 3:4 ] = NaN
S2
time Observations Predictions Pred_Variance
1 1 0.31296 NaN NaN
2 2 0.59448 NaN NaN
3 3 0.81627 0.95936 0.003832
4 4 0.95606 0.89728 0.011778
5 5 0.99978 0.88276 0.014781
6 6 0.94307 NaN NaN
7 7 0.79160 NaN NaN
8 8 0.56060 0.26670 0.076164
9 9 0.27328 0.28246 0.104256
10 10 -0.04149 0.36578 0.024815
11 11 NaN 0.09529 0.041300
12 12 NaN 0.17593 0.055783Since there is no difficulty predicting rows 6 & 7, as the library in this instance support it, is there a need to remove predictions from the output?