6.8.2. statsmodels.sandbox.stats.contrast_tools¶
functions to work with contrasts for multiple tests
contrast matrices for comparing all pairs, all levels to reference level, ... extension to 2-way groups in progress
TwoWay: class for bringing two-way analysis together and try out various helper functions
Idea for second part - get all transformation matrices to move in between different full rank
parameterizations
standardize to one parameterization to get all interesting effects.
multivariate normal distribution - exploit or expand what we have in LikelihoodResults, cov_params, f_test,
t_test, example: resols_dropf_full.cov_params(C2)
- connect to new multiple comparison for contrast matrices, based on multivariate normal or t distribution (Hothorn, Bretz, Westfall)
6.8.2.1. Functions¶
contrast_all_one(nm) |
contrast or restriction matrix for all against first comparison |
contrast_allpairs(nm) |
contrast or restriction matrix for all pairs of nm variables |
contrast_diff_mean(nm) |
contrast or restriction matrix for all against mean comparison |
contrast_labels(contrasts, names[, reverse]) |
|
contrast_product(names1, names2[, ...]) |
build contrast matrices for products of two categorical variables |
dummy_1d(x[, varname]) |
dummy variable for id integer groups |
dummy_limits(d) |
start and endpoints of groups in a sorted dummy variable array |
dummy_nested(d1, d2[, method]) |
unfinished and incomplete mainly copy past dummy_product |
dummy_product(d1, d2[, method]) |
dummy variable from product of two dummy variables |
groupmean_d(x, d) |
groupmeans using dummy variables |
signstr(x[, noplus]) |
6.8.2.2. Classes¶
DummyTransform(d1, d2) |
Conversion between full rank dummy encodings |
TestContrastTools() |
|
TwoWay(endog, factor1, factor2[, varnames]) |
a wrapper class for two way anova type of analysis with OLS |