Thank you for your suggestions. They are very helpful. I will use these techniques for my R scripts.

Based on your suggestions, I made the following examples.

First method to extend the data of dummy variables in "bitmets"

It is based on the use of TSJOIN() and TIMESERIES().

bimodel$modelData <- within(bimodel$modelData,{
dum_Q1_part2 <- TIMESERIES(0, 1, 0, 0, 0, 1, 0, 0, 0,
START = c(2022, 4), FREQ = 'Q')
dum_Q1 <- TSJOIN(dum_Q1, dum_Q1_part2, ALLOWGAP = FALSE)

dum_Q2_part2 <- TIMESERIES(0, 0, 1, 0, 0, 0, 1, 0, 0,
START = c(2022, 4), FREQ = 'Q')
dum_Q2 <- TSJOIN(dum_Q2, dum_Q2_part2, ALLOWGAP = FALSE)

dum_Q3_part2 <- TIMESERIES(0, 0, 0, 1, 0, 0, 0, 1, 0,
START = c(2022, 4), FREQ = 'Q')
dum_Q3 <- TSJOIN(dum_Q3, dum_Q3_part2, ALLOWGAP = FALSE)
})

end of the first method

Second method

p2_start <- c(2022, 4)
p2_end <- c(2024, 4)
p2_length <- NUMPERIOD(p2_start, p2_end, 4) + 1
zeroTS_p2 <- TSERIES(rep(0, p2_length), START = p2_start, FREQ = 'Q')
dum_Q1_p2 <- zeroTS_p2
dum_Q2_p2 <- zeroTS_p2
dum_Q3_p2 <- zeroTS_p2

quarterToBeOne <- 1
dum_Q1_p2[which(GETDATE(dum_Q1_p2, format = '%q') == quarterToBeOne)] <- 1

quarterToBeOne <- 2
dum_Q2_p2[which(GETDATE(dum_Q2_p2, format = '%q') == quarterToBeOne)] <- 1

quarterToBeOne <- 3
dum_Q3_p2[which(GETDATE(dum_Q3_p2, format = '%q') == quarterToBeOne)] <- 1

TABIT(dum_Q1_p2, dum_Q2_p2, dum_Q3_p2,
TSRANGE = c(2022, 4, 2024, 4))

bimodel$modelData <- within(bimodel$modelData,{
dum_Q1 <- TSJOIN(dum_Q1, dum_Q1_p2, ALLOWGAP = FALSE)
dum_Q2 <- TSJOIN(dum_Q2, dum_Q2_p2, ALLOWGAP = FALSE)
dum_Q3 <- TSJOIN(dum_Q3, dum_Q3_p2, ALLOWGAP = FALSE)
})

end of the second method

Check the content of new data set

TABIT(bimodel$modelData$G,
bimodel$modelData$MrP,
bimodel$modelData$TAX,
bimodel$modelData$dum_Q1,
bimodel$modelData$dum_Q2,
bimodel$modelData$dum_Q3,
TSRANGE = c(2021, 1, 2024, 4))

end of the examples

Note (G, MrP, TAX) are exogenous qualitative variables.

The data of (G, MrP, TAX) is extended with the usual method, using TSEXTEND() function.

from bimets.

dear @murao164jp

TSEXTEND doen not change current values in a time series. This function extends a time series outside its definition time range.

Lines 423:428 in forecast_out_sample_ISLM.TXT: In data.frame data_obs_V2, and in the related list data_preBi_V2, exogenous variables are already defined up tp 2024-4, and have missing values.

Therefore, when you apply TSEXTEND in lines 446, you are actually chaning nothing: time series are already defined up to 2024-4, and exogenous variables will still have missing values in simulation TSRANGE. Thus, SIMULATE issues an error.

If you want to replace missing values with arbitrary values in your time series, you have to manually remove missing values, or you may want to take a look at the TSTRIM function: after the missing removal, then you can apply the TSEXTEND.

from bimets.

Thank you for your comment.

I tried several ideas after posting my question, and I found a way to solve it.

It uses the combination of TSJOIN()function, TSEXTEND()function, TIMESERIES()function. My example is shown in the below.

A way to extend the data of dummy variables in "bimets"

The use of TSJOIN(), TSEXTEND(), and TIMESERIES() for it.

bimodel$modelData <- within(bimodel$modelData,{
dum_Q1_part2 = TSEXTEND(TIMESERIES(0, 1, 0, 0, 0, 1, 0, 0, 0, START = c(2022, 4), FREQ='Q'),
BACKTO = c(2022, 4), UPTO = c(2024, 4))
dum_Q1 <- TSJOIN(dum_Q1, dum_Q1_part2, ALLOWGAP = FALSE)

dum_Q2_part2 = TSEXTEND(TIMESERIES(0, 0, 1, 0, 0, 0, 1, 0, 0, START = c(2022, 4), FREQ='Q'),
BACKTO = c(2022, 4), UPTO = c(2024, 4))
dum_Q2 <- TSJOIN(dum_Q2, dum_Q2_part2, ALLOWGAP = FALSE)

dum_Q3_part2 = TSEXTEND(TIMESERIES(0, 0, 0, 1, 0, 0, 0, 1, 0, START = c(2022, 4), FREQ='Q'),
BACKTO = c(2022, 4), UPTO = c(2024, 4))
dum_Q3 <- TSJOIN(dum_Q3, dum_Q3_part2, ALLOWGAP = FALSE)

}) # end of within()

end of dealing with "bimodel$modelData"

Currently I have two suggestions.

(1) Please write your version of the above idea in the manual of bimets.

(2) It is nice to have this kind of option in TSEXTEND()function.

Thank you again.

Hiroshi Murao

from bimets.

hi @murao164jp

you may want to remove the TSEXTEND in your code, it does not change the target time series. In the following code, similar to yours, ts1 and ts1ext are the same time series:

> ts1=(TIMESERIES(0, 1, 0, 0, 0, 1, 0, 0, 0, START = c(2022, 4), FREQ='Q'))
> ts1ext=TSEXTEND(ts1,BACKTO = c(2022, 4), UPTO = c(2024, 4))
> TABIT(ts1,ts1ext)

      Date, Prd., ts1            , ts1ext

   2022 Q4, 4   ,  0             ,  0
   2023 Q1, 1   ,  1             ,  1
   2023 Q2, 2   ,  0             ,  0
   2023 Q3, 3   ,  0             ,  0
   2023 Q4, 4   ,  0             ,  0
   2024 Q1, 1   ,  1             ,  1
   2024 Q2, 2   ,  0             ,  0
   2024 Q3, 3   ,  0             ,  0
   2024 Q4, 4   ,  0             ,  0

If you want to create a time series that is one in an arbitrary quarter, you may want to take a look at the following:

> start=c(2022,4)
> end=c(2024,4)
> length=NUMPERIOD(start,end,4)+1
> tsBase=TSERIES(rep(0,length),START=start,FREQ='Q')

#set this variable to selected quarter, 1..4
> quarterToBeOne=3

> dummyQuarter3=tsBase
> dummyQuarter3[which(GETDATE(dummyQuarter3,format='%q')==quarterToBeOne)]=1
> TABIT(dummyQuarter3)

      Date, Prd., dummyQuarter3

   2022 Q4, 4   ,  0
   2023 Q1, 1   ,  0
   2023 Q2, 2   ,  0
   2023 Q3, 3   ,  1
   2023 Q4, 4   ,  0
   2024 Q1, 1   ,  0
   2024 Q2, 2   ,  0
   2024 Q3, 3   ,  1
   2024 Q4, 4   ,  0

Hope this will help.

from bimets.