Breaking changes:
Renamed the functions for downloading HICP data to
datasets(), datafilters() and
data().
Added the new functions aggregate.tree() and
disaggregate(). Moved the functionality of
aggregate() to aggregate.tree() to highlight
that this function is used for the gradual index aggregation following
the COICOP tree. The functions aggregate() and
disaggregate() can be used for the calculation of
user-defined aggregates.
Renamed the argument freq in convert()
to type, which is more in line with rates()
and contrib().
The function rates() now requires a date vector.
Before, the argument t was optional. This change makes the
calculation faster, more robust and better aligned to other package
functions.
Removed the list of COICOP bundles in
coicop.bundles, which are now added to the global package
options via options("hicp.coicop.bundles"). This allows
users to adjust the list of COICOP bundles.
The function tree() now returns a list with the full
COICOP tree at each level. For backward compatibility, the new argument
flag=TRUE can be used to still return a logical
vector.
New features and other changes:
The function contrib() now allows to derive
contributions to monthly, quarterly and annual change rates. This can
now also be done for quarterly and annual indices.
The function convert() now allows the calculation of
a 12-month moving average.
The function rebase() now allows for multiple index
reference periods in t.ref or setting this dynamically to
the first or last period.
The functions child() and parent() now
allow checking against the full COICOP code dictionary. Moreover, the
closest parents/children or the \(k\)-th parents/children can be
derived.
The function aggregate.tree() relies on
parent() and level(), which makes it more
robust but slightly slower.
Added settings$na.rm for dealing with
NAs when quarterly or yearly averages are computed. For
na.rm=TRUE, averages are now also calculated if there are
NAs or less than 12 observations.
Added settings$freq to set the frequency of the
index series manually. Otherwise, the frequency is
auto-detected.
Added settings$chatty for printing package specific
warnings and messages. This behavior can also be changed globally by
options("hicp.chatty").
Removed settings$unbundle. COICOP codes are now
always unbundled internally by the respective functions. If that is not
wanted by the user, the bundle codes can be set to NA or an
empty list can be defined in
options(hicp.coicop.bundles).
Years and months are now extracted using the functions
data.table::year() and data.table::month(),
which is much faster.
All functions using dates now check internally for duplicates to avoid daily or weekly data.
Added the argument pattern="^prc_hicp" to the
function datasets().
Added the codes of the new ECOICOP version 2.
Bug fixes:
Fixed a bug in the functions chain() and
unchain() when there was a gap of only one year and the
index value was set to 100 in December of this year. A break in the time
series is now assumed if the calendar year is missing/NA or
if only one index value is present in December (for
by=NULL) or in the month specified with
by.
Fixed a bug in child() for the derivation of
children.
r-release-windows-x86_64-flavorIntroduced the global options
options("hicp.unbundle"=TRUE),
options(hicp.coicop.version="ecoicop-hicp"),
options(hicp.all.items.code="00")
The functions is.coicop(), level(),
parent(), child(), aggregate(),
and tree() now have the argument settings,
which can be used to define a COICOP version
(coicop.version), the all-items code
(all.items.code), and how bundle codes are treated
(unbundle)
The function is.coicop() now accepts different
COICOP versions. The functions level(),
parent(), child(), aggregate(),
and tree() only process valid COICOP codes and the
all-items code.
Fixed the behavior of aggregate() when the functions
jevons(), carli(), and harmonic()
are used and no weights w0 are provided
Fixed a bug in the function chain(), which occurred
when the unchained series started in January and by=12. In
this case, the index reference period (with value 100) was not correctly
detected
Updated documentation and tests accordingly
Fixed a bug in the functions jevons(),
carli(), and harmonic(), which checked the
weights w0 that were, however, not required by the
function
Fixed a bug in the function chain(), which occurred
when NA’s were present in the index where the index value
was set to 100 (instead of NA)
Updated documentation and tests accordingly
New functions jevons(), carli(), and
harmonic() available
Function aggregate() now allows calculation of
user-defined aggregates even if not all parts of an aggregate are
present in the data (settings$add.exact=FALSE)
Function aggregate() no longer requires any weights
w0 or wt if the underlying index
function does not require weights
Function link() now returns a matrix instead of a
data.table
Function convert() now returns a named vector
instead of a data.table
Updated references to the new HICP Methodological Manual
Cleaned code for easier reference to other package functions
Updated description
Replaced \dontrun with \donttest in
examples
Initial CRAN submission.
First Github release.