Skip to content

← statistics package

MANCOVA

Calculate a (M)AN(C)OVA statistical model.

This node can be used to calculate ANOVA, MANOVA, ANCOVA, and MANCOVA solutions. The node will respect the statistical design described Describe Statistical Model (aka DescribeStatisticalDesign) node, or if not provided, will utilize all independent variables present in the data (specifically, in the instance axis of the data). For multivariate dependent variables, the node will always perform a mass-univariate analysis, as well as a multivariate analysis, which can however be optionally skipped, since it can be rather slow and is sometimes not possible if there are too few trials for the given number of dependent variables. The mass-univariate and multivariate ANOVAs are not guaranteed to use the same observations if there are scattered missing values in the dependent variables, which the MANOVA is robust to but the individual univariate ANOVAs are not. A common set of observations will be used for both univariate and multivariate analyses when the block data does not have any missing (NaN) values. The node outputs separately the statistics from the multivariate and univariate analysis, as well as the inferred statistical model (regression weights). The desired statistics can be selected subsequently using the Parse Stats Table node (or Select Range). The node supports parallel processing for additional speedup, as well as caching of the result (both may be disabled by default, but can provide tremendous speed gains, e.g., when repeatedly re-running the same analysis). This node creates three new axes: a Statistic axis with the label 'statistic_types' containing the stats (i.e. F-value), a Statistic axis with the label 'multivar_test_types' containing the multivariate test types (i.e., Wilks lambda), and a Feature axis with the label 'explanatory_variables' containing the design terms (i.e., intercept). (Note: For backward compatibility reasons (versions < 2.2.0), legacy_mode is set to True and therefore this node will store all stats in Feature axes axis instead of Statistic axes. Please update your pipelines as needed to assume a Statistic axis and then set legacy_mode to False. This will be deprecated in a future release.) Version 2.2.2

Ports/Properties

data

A packet containing at least one signal chunk with an instance axis. The chunk may have gone through DescribeStatisticalDesign previously to help design the statistical test when it cannot be inferred easily by inspecting the instance axis data.

verbose name
Data
default value
None
port type
DataPort
value type
Packet (can be None)
data direction
INOUT

multivar_stats

For each data chunk with an instance axis, this will output a 3D tensor that has, for each explanatory variable, the multivariate test type, the statistic type, and the value of that statistic

verbose name
Multivar Stats
default value
None
port type
DataPort
value type
object (can be None)
data direction
OUT

univar_stats

Univariate post-hoc statistics.

verbose name
Univar Stats
default value
None
port type
DataPort
value type
Packet (can be None)
data direction
OUT

result

The MANCOVA model.

verbose name
Result
default value
None
port type
DataPort
value type
object (can be None)
data direction
OUT

univar_teststat

Test statistic for univariate ANCOVA tests. These post-hoc tests are used to identify which of the response variables showed a significant effect.

verbose name
Univariate Test Statistic
default value
F
port type
EnumPort
value type
str (can be None)

univar_testtype

Test type to use for univariate ANCOVA tests. These post-hoc tests are used to identify which of the response variables showed a significant effect.

verbose name
Univariate Test Type
default value
I
port type
EnumPort
value type
str (can be None)

univar_robust

Heteroscedasticity correction for robust covariance analysis. Used for univariate post-hoc tests.

verbose name
Univariate Hc Correction
default value
None
port type
EnumPort
value type
str (can be None)

skip_mvar

Skip multivariate statistics and return only univariate results.

verbose name
Skip Multivariate Analysis
default value
False
port type
BoolPort
value type
bool (can be None)

missing

Choose how to handle missing values. If 'drop', any observations with nans in the exogenous IVs are dropped. MANOVA can tolerate some missing values in the multivariate endogenous DV so these are not dropped. Unlike the MANOVA, the mass univariate ANOVAs will drop any observations with nan in the univariate DV on a per-variable basis. If 'raise', an error is raised. Default is 'drop'.

verbose name
Handle Missing Data
default value
drop
port type
EnumPort
value type
str (can be None)

par_cores

Number of cores to use in parallel processing. 1: Do not do parallel processing. -1: Auto-detect the number of cores. >1: Do parallel processing with this many cores.

verbose name
Number Of Cores To Use
default value
1
port type
IntPort
value type
int (can be None)

legacy_mode

Store the statistics in a Feature axis instead of a Statistic axis. Included for backward compatibility and will be deprecated.

verbose name
Legacy Output Format
default value
True
port type
BoolPort
value type
bool (can be None)

use_caching

Enable caching.

verbose name
Use Caching
default value
False
port type
BoolPort
value type
bool (can be None)

verbose

Verbose output. Can be used for debugging your pipeline.

verbose name
Verbose
default value
False
port type
BoolPort
value type
bool (can be None)

set_breakpoint

Set a breakpoint on this node. If this is enabled, your debugger (if one is attached) will trigger a breakpoint.

verbose name
Set Breakpoint (Debug Only)
default value
False
port type
BoolPort
value type
bool (can be None)

metadata

User-definable meta-data associated with the node. Usually reserved for technical purposes.

verbose name
Metadata
default value
{}
port type
DictPort
value type
dict (can be None)