Function generates the matrix of dummy variables for the months / weeks / days / hours / minutes / seconds of year / month / week / day / hour / minute.
temporaldummy(object, type = c("month", "quarter", "week", "day", "hour",
"halfhour", "minute", "second"), of = c("year", "quarter", "month", "week",
"day", "hour", "minute"), h = 0)# S3 method for default
temporaldummy(object, type = c("month", "quarter", "week",
"day", "hour", "halfhour", "minute", "second"), of = c("year", "quarter",
"month", "week", "day", "hour", "minute"), h = 0)
Either a ts / msts / zoo / xts / tsibble object or a vector of dates.
Specifies what type of frequency to produce. For example, if
type="month", then the matrix with dummies for months of the year will
be created.
Specifies the frequency of what is needed. Used together with type
e.g. type="day" and of="month" will produce a matrix with dummies
for days of month (31 dummies).
If not NULL, then the function will produce dummies for this
set of observations ahead as well, binding them to the original matrix.
Class "dgCMatrix" with all the dummy variables is returned in case of numeric variable. Feel free to drop one (making it a reference variable) or convert the object into matrix (this will consume more memory than the returned class). In other cases the object of the same class as the provided is returned.
The function extracts dates from the provided object and returns a matrix with dummy variables for the specified frequency type, with the number of rows equal to the length of the object + the specified horizon. If a numeric vector is provided then it will produce dummies based on typical values (e.g. 30 days in month). So it is recommended to use proper classes with this method.
Several notes on how the dummies are calculated in some special cases:
In case of weeks of years, the first week is defined according to ISO 8601.
Note that not all the combinations of type and of are supported. For
example, there is no such thing as dummies for months of week. Also note that some
combinations are not very useful and would take a lot of memory (e.g. minutes of year).
The function will return all the dummy variables. If you want to avoid the dummy variables trap, you will need to exclude one of them manually.
If you want to have a different type of dummy variables, let me know, I will implement it.
# NOT RUN {
# Generate matrix with dummies for a ts object
x <- ts(rnorm(100,100,1),frequency=12)
temporaldummy(x)
# Generate matrix with monthly dummies for a zoo object
x <- as.Date("2003-01-01")+0:99
temporaldummy(x, type="month", of="year", h=10)
# }
Run the code above in your browser using DataLab