Y = cumsumk(X,K,Rho,Range)
Y = cumsumk(X,K,Rho)
Y = cumsumk(X,K)
Y = cumsumk(X)
X
[ tseries ] - Input data.
K
[ numeric ] - Number of periods that will be leapt the cumulative sum will be taken; if not specified, K
is chosen to match the frequency of the input data (e.g. K = -4
for quarterly data), or K = -1
for indeterminate frequency.
Rho
[ numeric ] - Autoregressive coefficient; if not specified, Rho = 1
.
Range
[ numeric ] - Range on which the cumulative sum will be computed and the output series returned.
Y
[ tseries ] - Output data constructed as described below.'log='
[ true
| false
] - Logarithmise the input data before, and de-logarithmise the output data back after, running x12
.If K < 0
, the first K
observations in the output series Y
are copied from X
, and the new observations are given recursively by
Y{t} = Rho*Y{t-K} + X{t}.
If K > 0
, the last K
observations in the output series Y
are copied from X
, and the new observations are given recursively by
Y{t} = Rho*Y{t+K} + X{t},
going backwards in time.
If K == 0
, the input data are returned.
Construct random data with seasonal pattern, and run X12 to seasonally adjust these series.
x = tseries(qq(1990,1):qq(2020,4),@randn);
x1 = cumsumk(x,-4,1);
x2 = cumsumk(x,-4,0.7);
x1sa = x12(x1);
x2sa = x12(x2);
The new series x1
will be a unit-root process while x2
will be stationary. Note that the command on the second line could be replaced with x1 = cumsumk(x)
.