Datasets¶

fdatools.datasets: the datasets used in the FDA books. growth, gait and pinch ship inside the package. The others are downloaded once, checked against a SHA-256 checksum and cached in ~/.cache/fdatools (set FDATOOLS_DATA_DIR to change the location).

Loaders for the real datasets shipped with R's fda package (v6.3.0).

Every load_*() function returns a frozen dataclass of raw NumPy arrays plus the string/label metadata R attaches to the object (station names, subject ids, factor levels, ...), reproduced exactly as fda stores them. None of these datasets is already a smooth functional object in R (fda stores them as plain matrices/arrays/data frames), so none of the dataclasses below carries an .fd property -- fit one with :func:fdatools.smooth using the raw arrays and the accompanying t grid.

Storage: growth, gait and pinch ship inside the wheel (src/fdatools/_data/, read via :mod:importlib.resources). Every other dataset is downloaded on first use from the project's GitHub release data-v1, SHA-256 verified, and cached under $FDATOOLS_DATA_DIR (default ~/.cache/fdatools). See docs/dev/data-release.md for how the release archives are built and published.

fda 6.3.0's dataset catalog has no CSTR dataset and no separate "Chinese script" dataset (see tests/golden/datasets.json's dataset_catalog case) -- load_cstr/load_chinese_script from SPEC.md §8 are not implemented; see PROGRESS.md for the record. The catalog's "dateAccessories" group (day.5, monthLetters, ...) are plotting-axis constants, not datasets, and are likewise out of scope.

CanadianWeather `dataclass` ¶

CanadianWeather(
    temp: NDArray[float64],
    precip: NDArray[float64],
    log10precip: NDArray[float64],
    t: NDArray[float64],
    stations: list[str],
    province: list[str],
    region: list[str],
    coordinates: NDArray[float64],
    monthly_temp: NDArray[float64],
    monthly_precip: NDArray[float64],
    geogindex: NDArray[float64],
)

Daily temperature/precipitation at 35 Canadian weather stations, 1960-1994.

Reproduces R fda::CanadianWeather (dailyAv's three layers split into separate fields) plus CanadianWeather$monthlyTemp/monthlyPrecip.

Attributes:

Name	Type	Description
`temp, precip`	`(ndarray, shape(365, 35))`	Average daily temperature (deg C) and precipitation (mm).
`log10precip`	`(ndarray, shape(365, 35))`	`log10(precip)` with the 27 zero readings replaced by 0.05 mm first.
`t`	`(ndarray, shape(365))`	Day of year, 1-365.
`stations`	`list of str`	35 station names (R: `CanadianWeather$place`).
`province, region`	`list of str`	Province and one of the 4 climate zones (Atlantic/Pacific/Continental/Arctic) for each station.
`coordinates`	`(ndarray, shape(35, 2))`	North latitude, west longitude.
`monthly_temp, monthly_precip`	`(ndarray, shape(12, 35))`	Monthly averages: temperature (deg C) and daily precipitation (mm).
`geogindex`	`(ndarray, shape(35))`	Station order east-to-west-then-north (R: `CanadianWeather$geogindex`).

Growth `dataclass` ¶

Growth(
    hgtm: NDArray[float64],
    hgtf: NDArray[float64],
    age: NDArray[float64],
)

Berkeley Growth Study: heights of 39 boys and 54 girls, ages 1-18.

Reproduces R fda::growth (hgtm, hgtf, age). Ages are not equally spaced.

Attributes:

Name	Type	Description
`hgtm`	`(ndarray, shape(31, 39))`	Boys' heights in cm at each of 31 ages.
`hgtf`	`(ndarray, shape(31, 54))`	Girls' heights in cm at each of 31 ages.
`age`	`(ndarray, shape(31))`	Ages in years at which heights were measured.

Gait `dataclass` ¶

Gait(
    value: NDArray[float64],
    t: NDArray[float64],
    subjects: list[str],
    variables: list[str],
)

Hip and knee angle (degrees) through a 20-point gait cycle, 39 boys.

Reproduces R fda::gait, an array of dim (20, 39, 2).

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(20, 39, 2))`	Angles in degrees: `value[..., 0]` is hip angle, `value[..., 1]` is knee angle (see :attr:`hip_angle`/:attr:`knee_angle`).
`t`	`(ndarray, shape(20))`	Standardized gait time, proportion of the cycle (0.025 to 0.975).
`subjects`	`list of str`	39 subject ids, "boy1".."boy39".
`variables`	`list of str`	`["Hip Angle", "Knee Angle"]`, the meaning of `value`'s last axis.

hip_angle `property` ¶

hip_angle: NDArray[float64]

Return value[..., 0], shape (20, 39).

knee_angle `property` ¶

knee_angle: NDArray[float64]

Return value[..., 1], shape (20, 39).

Handwriting `dataclass` ¶

Handwriting(
    value: NDArray[float64],
    t: NDArray[float64],
    subjects: list[str],
    variables: list[str],
)

20 cursive samples of 1401 (x, y) coordinates writing "fda".

Reproduces R fda::handwrit + fda::handwritTime. The subject was Jim Ramsay; each replicate has been resampled to a common 2300 ms length and registered so key features align across replicates.

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(1401, 20, 2))`	Pen position in metres: `value[..., 0]` is X, `value[..., 1]` is Y (see :attr:`x`/:attr:`y`).
`t`	`(ndarray, shape(1401))`	Sampling times in milliseconds, 0 to 2300.
`subjects`	`list of str`	20 replicate ids, "rep01".."rep20".
`variables`	`list of str`	`["X", "Y"]`, the meaning of `value`'s last axis.

x `property` ¶

x: NDArray[float64]

Return value[..., 0], shape (1401, 20).

y `property` ¶

y: NDArray[float64]

Return value[..., 1], shape (1401, 20).

Pinch `dataclass` ¶

Pinch(
    pinch: NDArray[float64],
    pinchraw: NDArray[float64],
    t: NDArray[float64],
)

151 measurements of pinch force across 20 replications.

Reproduces R fda::pinch + pinchraw + pinchtime. The original recording had 300 observations every 2 ms; pinchraw keeps the first 151, while pinch instead selects the 151 observations so that every curve's maximum falls at 0.076 s.

Attributes:

Name	Type	Description
`pinch`	`(ndarray, shape(151, 20))`	Pinch force in newtons (N), aligned so each curve peaks at 0.076 s.
`pinchraw`	`(ndarray, shape(151, 20))`	Pinch force in newtons (N), the first 151 of the original 300 samples, unaligned.
`t`	`(ndarray, shape(151))`	Time in seconds from the start, every 2 ms (R: `pinchtime`).

Melanoma `dataclass` ¶

Melanoma(value: NDArray[float64], columns: list[str])

Age-adjusted melanoma incidence per 100,000 in Connecticut, 1936-1972.

Reproduces R fda::melanoma. The R object is a bare 37x3 matrix; its first column is a plain 1..37 row index (the documented format lists only year/incidence, but the shipped object carries a leading index column -- reproduced as-is, see columns for the mapping).

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(37, 3))`	Columns as listed in `columns`: the row index, the calendar year (1936-1972), and the age-adjusted incidence per 100,000 people.
`columns`	`list of str`	`["index", "year", "incidence"]`.

Refinery `dataclass` ¶

Refinery(
    time: NDArray[float64],
    reflux: NDArray[float64],
    tray47: NDArray[float64],
)

194 observations of reflux and "tray 47 level" in an oil refinery column.

Reproduces R fda::refinery (a data frame with 3 numeric columns). R documents no physical unit for any column, so none is given here.

Attributes:

Name	Type	Description
`time`	`(ndarray, shape(194))`	Observation time, 0-193 (an index; R gives no unit).
`reflux`	`(ndarray, shape(194))`	Reflux flow, centered on the mean of the first 60 observations (no unit documented).
`tray47`	`(ndarray, shape(194))`	Tray 47 level, centered on the mean of the first 60 observations (no unit documented).

Seabird `dataclass` ¶

Seabird(
    counts: dict[str, NDArray[float64]],
    year: NDArray[float64],
    site: NDArray[float64],
    transect: NDArray[float64],
    temp: NDArray[float64],
    observ_cond: list[str],
    bay: list[str],
    observ_cond_factor3: list[str] = list(),
)

Seabird sighting counts by transect, 4 Kodiak Island bays, 1986-2005.

Reproduces R fda::seabird (a data frame with 22 columns): 15 species count columns, 3 integer id columns, one numeric column, and 3 factor columns.

Attributes:

Name	Type	Description
`counts`	`dict of str to ndarray, each shape (3793,)`	One entry per species code (`BAGO`, `BLSC`, `COME`, `COMU`, `CORM`, `HADU`, `HOGR`, `LOON`, `MAMU`, `OLDS`, `PIGU`, `RBME`, `RNGR`, `SUSC`, `WWSC`): the number of sightings of that species by transect by year; `NaN` marks R's `NA`.
`year, site, transect, temp`	`(ndarray, shape(3793))`	Survey year (1986-2005), site code, transect code, and temperature (R documents no unit for the temperature).
`observ_cond`	`list of str`	Observing conditions factor (5 levels).
`bay`	`list of str`	Bay name factor (4 levels).
`observ_cond_factor3`	`list of str`	`observ_cond` collapsed to 3 levels.

ReginaPrecip `dataclass` ¶

ReginaPrecip(value: NDArray[float64])

Daily June precipitation (mm) in Regina, Saskatchewan, 1960-1993.

Reproduces R fda::ReginaPrecip: 16 missing values already omitted.

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(1006))`	Precipitation in mm.

MontrealTemp `dataclass` ¶

MontrealTemp(value: NDArray[float64], years: list[str])

Daily temperature (deg C) in Montreal, 1961-1994.

Reproduces R fda::MontrealTemp, a 34 (years) x 365 (days) matrix.

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(34, 365))`	Daily average temperature in degrees Celsius.
`years`	`list of str`	The 34 years, "1961".."1994".

Daily `dataclass` ¶

Daily(
    tempav: NDArray[float64],
    precav: NDArray[float64],
    place: list[str],
)

Daily temperature/precipitation at 35 Canadian stations (alternate layout).

Reproduces R fda::daily: the same data as :class:CanadianWeather, kept by fda "primarily for compatibility with scripts written before the other format became available" and with the Matlab fda code -- place here is padded to a fixed 11 characters (unlike :attr:CanadianWeather.stations, which is trimmed).

Attributes:

Name	Type	Description
`tempav, precav`	`(ndarray, shape(365, 35))`	Average daily temperature (deg C) and precipitation (mm).
`place`	`list of str`	35 station names, each padded to 11 characters with trailing blanks.

InfantGrowth `dataclass` ¶

InfantGrowth(
    day: NDArray[float64],
    tibia_length: NDArray[float64],
    sd_length: NDArray[float64],
)

Tibia length for one infant's first 40 days of life.

Reproduces R fda::infantGrowth (a 3-column matrix): each length is the average of 5 measurements.

Attributes:

Name	Type	Description
`day`	`(ndarray, shape(40))`	Age in days.
`tibia_length`	`(ndarray, shape(40))`	Average tibia length in mm.
`sd_length`	`(ndarray, shape(40))`	Standard deviation of the 5 measurements, in mm.

Nondurables `dataclass` ¶

Nondurables(
    value: NDArray[float64], start: str, frequency: int
)

US nondurable goods index, monthly, January 1919 onward.

Reproduces R fda::nondurables.

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(1377))`	The index value each month (a dimensionless index; R documents no unit).
`start`	`str`	First month covered, `"1919-01"`.
`frequency`	`int`	Observations per year (12, monthly).

Lip `dataclass` ¶

Lip(
    value: NDArray[float64],
    t: NDArray[float64],
    left_elbow: NDArray[float64],
    right_elbow: NDArray[float64],
)

51 measurements of lower lip position, 20 repetitions of "bob".

Reproduces R fda::lip + liptime + lipmarks.

Attributes:

Name	Type	Description
`value`	`(ndarray, shape(51, 20))`	Lower lip position in mm (the unit R's help examples give it), sampled every 7 ms for 350 ms.
`t`	`(ndarray, shape(51))`	Time in seconds from the start (R: `liptime`).
`left_elbow, right_elbow`	`(ndarray, shape(20))`	Landmark times (R: `lipmarks$leftElbow`/`lipmarks$rightElbow`) for each of the 20 repetitions.

load_canadian_weather ¶

load_canadian_weather() -> CanadianWeather

Load :class:CanadianWeather (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> cw = fdt.datasets.load_canadian_weather()
>>> cw.temp.shape
(365, 35)

load_growth ¶

load_growth() -> Growth

Load :class:Growth (ships in-package, no network).

Examples:

>>> import fdatools as fdt
>>> g = fdt.datasets.load_growth()
>>> g.hgtm.shape, g.hgtf.shape, g.age.shape
((31, 39), (31, 54), (31,))

load_gait ¶

load_gait() -> Gait

Load :class:Gait (ships in-package, no network).

Examples:

>>> import fdatools as fdt
>>> gait = fdt.datasets.load_gait()
>>> gait.value.shape
(20, 39, 2)

load_handwriting ¶

load_handwriting() -> Handwriting

Load :class:Handwriting (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> hw = fdt.datasets.load_handwriting()
>>> hw.value.shape
(1401, 20, 2)

load_pinch ¶

load_pinch() -> Pinch

Load :class:Pinch (ships in-package, no network).

Examples:

>>> import fdatools as fdt
>>> p = fdt.datasets.load_pinch()
>>> p.pinch.shape, p.pinchraw.shape, p.t.shape
((151, 20), (151, 20), (151,))

load_melanoma ¶

load_melanoma() -> Melanoma

Load :class:Melanoma (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> m = fdt.datasets.load_melanoma()
>>> m.value.shape
(37, 3)

load_refinery ¶

load_refinery() -> Refinery

Load :class:Refinery (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> r = fdt.datasets.load_refinery()
>>> r.time.shape
(194,)

load_seabird ¶

load_seabird() -> Seabird

Load :class:Seabird (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> sb = fdt.datasets.load_seabird()
>>> sb.year.shape
(3793,)

load_regina_precip ¶

load_regina_precip() -> ReginaPrecip

Load :class:ReginaPrecip (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> rp = fdt.datasets.load_regina_precip()
>>> rp.value.shape
(1006,)

load_montreal_temp ¶

load_montreal_temp() -> MontrealTemp

Load :class:MontrealTemp (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> mt = fdt.datasets.load_montreal_temp()
>>> mt.value.shape
(34, 365)

load_daily ¶

load_daily() -> Daily

Load :class:Daily (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> d = fdt.datasets.load_daily()
>>> d.tempav.shape
(365, 35)

load_infant_growth ¶

load_infant_growth() -> InfantGrowth

Load :class:InfantGrowth (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> ig = fdt.datasets.load_infant_growth()
>>> ig.day.shape
(40,)

load_nondurables ¶

load_nondurables() -> Nondurables

Load :class:Nondurables (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> nd = fdt.datasets.load_nondurables()
>>> nd.value.shape
(1377,)

load_lip ¶

load_lip() -> Lip

Load :class:Lip (downloaded + cached on first use).

Examples:

>>> import fdatools as fdt
>>> lip = fdt.datasets.load_lip()
>>> lip.value.shape
(51, 20)

Datasets¶

CanadianWeather dataclass ¶

Growth dataclass ¶

Gait dataclass ¶

hip_angle property ¶

knee_angle property ¶

Handwriting dataclass ¶

x property ¶

y property ¶

Pinch dataclass ¶

Melanoma dataclass ¶

Refinery dataclass ¶

Seabird dataclass ¶

ReginaPrecip dataclass ¶

MontrealTemp dataclass ¶

Daily dataclass ¶

InfantGrowth dataclass ¶

Nondurables dataclass ¶

Lip dataclass ¶

load_canadian_weather ¶

load_growth ¶

load_gait ¶

load_handwriting ¶

load_pinch ¶

load_melanoma ¶

load_refinery ¶

load_seabird ¶

load_regina_precip ¶

load_montreal_temp ¶

load_daily ¶

load_infant_growth ¶

load_nondurables ¶

load_lip ¶

CanadianWeather `dataclass` ¶

Growth `dataclass` ¶

Gait `dataclass` ¶

hip_angle `property` ¶

knee_angle `property` ¶

Handwriting `dataclass` ¶

x `property` ¶

y `property` ¶

Pinch `dataclass` ¶

Melanoma `dataclass` ¶

Refinery `dataclass` ¶

Seabird `dataclass` ¶

ReginaPrecip `dataclass` ¶

MontrealTemp `dataclass` ¶

Daily `dataclass` ¶

InfantGrowth `dataclass` ¶

Nondurables `dataclass` ¶

Lip `dataclass` ¶