Difference between decomposition and finals

[TanguyBarthelemy]

In the output of x13 function some series are duplicated between decomposition and finals.

For example:

x <- AirPassengers +
    c(rep(0, 100), 200, rep(0, 43)) + # Ajout d'un AO en mai 1957
    c(rep(0, 80), rep(-100, 64)) # Ajout d'un LS en septembre 1955
plot(x)

^{Created on 2024-10-04 with reprex v2.1.1}

d11

There are 2 variables decomposition.d11 and finals.d11.

mod_v3 <- rjd3x13::x13(x, spec = "RSA3", userdefined = c("decomposition.d11", "finals.d11"))
d11 <- cbind(mod_v3$user_defined$finals.d11, mod_v3$user_defined$decomposition.d11)
ts.plot(d11, col = c("blue", "red"), lwd = c(2, 1))
legend("left", c("finals.d11", "decomposition.d11"), col = c("blue", "red"), lwd = c(2, 1))

^{Created on 2024-10-04 with reprex v2.1.1}

d12

Same thing for d12

mod_v3 <- rjd3x13::x13(x, spec = "RSA3", userdefined = c("decomposition.d12", "finals.d12"))
d12 <- cbind(mod_v3$user_defined$finals.d12, mod_v3$user_defined$decomposition.d12)
ts.plot(d12, col = c("blue", "red"), lwd = c(2, 1))
legend("left", c("finals.d12", "decomposition.d12"), col = c("blue", "red"), lwd = c(2, 1))

^{Created on 2024-10-04 with reprex v2.1.1}

d13

Same thing for d13

mod_v3 <- rjd3x13::x13(x, spec = "RSA3", userdefined = c("decomposition.d13", "finals.d13"))
d13 <- cbind(mod_v3$user_defined$finals.d13, mod_v3$user_defined$decomposition.d13)
ts.plot(d13, col = c("blue", "red"), lwd = c(2, 1))
legend("left", c("finals.d13", "decomposition.d13"), col = c("blue", "red"), lwd = c(2, 1))

^{Created on 2024-10-04 with reprex v2.1.1}

It seems that the d series in decomposition are corrected from the outliers (linearised series).
Should this series be renamed?

[AQLT]

Yes this is not as in V2 where the final components of the decomposition table also includes the pre-processing. In V3 you clearly separate the pre-processing, the decomposition and the final components (finals). This allow to more easily compute some statistics (for example the residual seasonality tests are made on the decomposition table and not the final tables). For example, those components where recently added in RJDemetra (rjdverse/rjdemetra#139)

[TanguyBarthelemy]

Ok yes, but I add in mind that the d11 component referred to a single series. So why are finals.d11 and decomposition.d11 different? Shouldn't we be talking about linearised series instead or in the X11 process, isn't there another name for these series with or without outliers?

[annasmyk]

we discussed this, I close it for now