Skip to content

Statistics: cicada_tools.stats

from cicada_analysis.cicada_tools.stats.<module> import <function>


stats.stats (Statistical tests, lines and bands of groups)

  • compare_two_distributions

compare_two_distributions(distribution1, distribution2, pvalues=0.05, verbose=True) -> None

Compare two distributions with several two-sample tests and print the results (nothing returned).

Parameter Type Default Description
distribution1 any First distribution
distribution2 any Second distribution
pvalues float 0.05 Significance threshold
verbose bool True Print results

Returns: None

  • multiple_comparison_one_factor_effect

multiple_comparison_one_factor_effect(distributions_list, pvalues=0.05, verbose=True, sessions_ids=None) -> None

Non-parametric one-factor test across multiple groups (Kruskal-Wallis + post-hoc).

Parameter Type Default Description
distributions_list list List of distributions
pvalues float 0.05 Significance threshold
verbose bool True Print results
sessions_ids list None Session IDs

Returns: None

  • compare_groups

compare_groups(rows, value, x, hue=None, pvalues=0.05, verbose=True, units=None) -> None

Compares the distributions of value between the groups of column x within each group of column hue (two groups: compare_two_distributions, more: multiple_comparison_one_factor_effect; NaN left out).

Parameter Type Default Description
rows pd.DataFrame Data rows
value str Value column
x str Group column
hue str None Hue column
pvalues float 0.05 Significance threshold
verbose bool True Print results
units list[str] None Columns of the statistical unit (e.g. ['mouse']): the mean of value per unit and group is compared, so n is the number of units. None: every row is a unit

Returns: None

  • summarize_by_levels

summarize_by_levels(table, value, group_columns, levels=...) -> dict

Count, mean, SD, min, median and max of value per combination of group_columns, per session and per mouse by default: {level name: summary table}.

Parameter Type Default Description
table pd.DataFrame Input table
value str Value column
group_columns list Group columns
levels any Levels

Returns: dict — Summary tables by level

  • chi_square_uniformity_test

chi_square_uniformity_test(observed_data, binsize) -> tuple

Chi-square goodness-of-fit test of the histogram of observed_data (bins of binsize) against a uniform distribution. Returns (chi2_stat, p_value, uniformity) (uniformity: p ≥ 0.05).

Parameter Type Description
observed_data any Observed data
binsize float Bin size

Returns: tuple — (chi2_stat, p_value, uniformity)

  • line_and_band

line_and_band(data, value, by, reduction='median', band='percentiles', low_percentile=25, high_percentile=75) -> pd.DataFrame

Line and band of a value over the rows of each group, e.g. a PSTH at each time point. NaN values are ignored. Band: 'SD' / 'SEM' = line ± SD / SEM (ddof 1), '95% CI' = mean ± the bootstrap half-width of the mean (slow), 'percentiles' = low_percentile to high_percentile, 'none' = no band. No band for a group with fewer than 2 values (as seaborn).

Parameter Type Default Description
data pd.DataFrame One row per unit of the level reduced (event, session, mouse...)
value str Column reduced
by list[str] Columns of the groups (e.g. ['bodypart', 'trial_name', 'time'])
reduction str 'median' 'median' or 'mean'
band str 'percentiles' 'SD', 'SEM', '95% CI', 'percentiles' or 'none'
low_percentile float 25 Lower limit of the 'percentiles' band
high_percentile float 75 Upper limit of the 'percentiles' band

Returns: pd.DataFrame — one row per group, in order of appearance: the by columns, line, low, high (NaN without band) and n (number of non-NaN values)

  • band_functions

band_functions(reduction='median', band='percentiles', low_percentile=25, high_percentile=75) -> list

The line and band of line_and_band as functions of a 1-D array, NaN ignored, for code that reduces one time point at a time (e.g. fcts_to_apply of build_population_psth): no line without values, no band under 2 values.

Parameter Type Default Description
reduction str 'median' As in line_and_band
band str 'percentiles' As in line_and_band
low_percentile float 25 As in line_and_band
high_percentile float 75 As in line_and_band

Returns: list — [line, low, high], three functions of a 1-D array returning a float

  • array_line_and_band

array_line_and_band(values, axis=0, reduction='median', band='percentiles', low_percentile=25, high_percentile=75) -> (line, low, high)

The line and band of line_and_band along one axis of an array (e.g. the trials of each unit and time bin): NaN ignored, no band under 2 values. A band other than the five of line_and_band gives no band.

Parameter Type Default Description
values np.ndarray Values, reduced along axis
axis int 0 Axis reduced
reduction str 'median' As in line_and_band
band str 'percentiles' As in line_and_band
low_percentile float 25 As in line_and_band
high_percentile float 75 As in line_and_band

Returns: tuple — (line, low, high), each with the shape of values without axis

  • reduce_psth_level

reduce_psth_level(rows, by, reduction='median', band='percentiles', low_percentile=25, high_percentile=75) -> pd.DataFrame

Reduce PSTH lines one level up (e.g. the sessions of a mouse, or the mice of a group): line and band of the rows' line values in each group of by (line_and_band). A group made of a single line keeps the band it had at the level below (e.g. a single session keeps its band over events).

Parameter Type Default Description
rows pd.DataFrame One row per unit and time point, with line, low, high and the by columns
by list[str] Columns of the groups, the time column included
reduction str 'median' As in line_and_band
band str 'percentiles' As in line_and_band
low_percentile float 25 As in line_and_band
high_percentile float 75 As in line_and_band

Returns: pd.DataFrame — the by columns, line, low, high, n (number of lines reduced)