Capability table

Options and evidence for the flagship estimators, read from their signatures and the registry. Evidence tier is the function-level summary (sp.describe_function(name)['validation_status']); config map says whether sp.validation_scope can tell you which configurations and outputs of a fit were actually compared. A tier never covers options that no evidence row ran -- check the map (or sp.result_card(fit)['evidence']) for the call you made. Support tier (sp.support_tier(name)) is derived: core = stable with numerical evidence, extension = stable and contract-tested without it, research = experimental or heuristic (LLM / neural).

Linear / GLM

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.regress	weights	cluster	vce=None, vcov=None, robust='nonrobust'	native	core	certified	yes	—
sp.logit	weights	cluster	robust='nonrobust'	native	core	certified	yes	—
sp.probit	weights	cluster	robust='nonrobust'	native	core	certified	yes	—
sp.poisson	weights	cluster	robust='nonrobust'	native	core	certified	yes	—
sp.nbreg	weights	cluster	robust='nonrobust'	native	core	certified	yes	—
sp.glm	weights	cluster	robust='nonrobust'	native	core	certified	yes	—

Panel / HDFE

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.feols	weights	cluster	vcov=None	pyfixest (varying slopes: native)	core	certified	yes	—
sp.hdfe_ols	weights	cluster	vce=None, se_type=None	native	core	certified	—	—
sp.panel	via **kwargs	via **kwargs	via **kwargs	native	core	certified	yes	—
sp.fepois	weights	cluster	vcov=None	pyfixest	core	certified	—	—

IV

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.iv	via **kwargs	via **kwargs	via **kwargs	native	core	certified	yes	—
sp.liml	—	cluster	robust='nonrobust'	native	core	certified	—	—

DiD / event study

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.did	weights	cluster	vce=None, robust=True, se_method='placebo'	native	core	validated	—	—
sp.callaway_santanna	weights	clustervars	se_method=None	native	core	certified	yes	1
sp.sun_abraham	weights	cluster	—	native	core	certified	yes	—
sp.did_imputation	weights	cluster	vce='analytic', se_method=None	native	core	certified	yes	1
sp.etwfe	weights	cluster	—	native	core	certified	yes	8
sp.did_multiplegt_dyn	weights	cluster	se_method=None	native	research	experimental	yes	3
sp.honest_did	—	—	—	native (backend='native')	core	certified	yes	—

RD

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.rdrobust	weights	cluster	vce='nn'	native	core	certified	yes	2
sp.rddensity	—	—	—	native (backend='native')	core	certified	yes	1

Synthetic control

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.synth	via **kwargs	via **kwargs	via **kwargs	native	core	certified	yes	2
sp.sdid	—	—	se_method='placebo'	native (backend='native')	core	certified	yes	—

Selection on observables / DML

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.psm	weights	via **kwargs	via **kwargs	native	core	certified	yes	—
sp.match	via **kwargs	via **kwargs	via **kwargs	native	core	certified	—	2
sp.ipw	weights	cluster	se_method='bootstrap'	native	core	certified	yes	—
sp.aipw	weights	cluster	se_method='influence'	native	core	certified	yes	—
sp.tmle	weights	cluster	se_method='influence'	native	core	certified	yes	—
sp.dml	via **kwargs	via **kwargs	via **kwargs	native	core	certified	yes	3
sp.causal_forest	via **kwargs	clusters	via **kwargs	native	core	certified	yes	4

Survey / MI

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.svymean	—	—	—	native	core	certified	—	—
sp.svyglm	—	—	—	native	core	certified	—	—
sp.mice	—	—	—	native	core	validated	—	—

Postestimation

function	weights	clustering	variance options	backend	support tier	evidence tier	config map	limitations
sp.margins	—	—	vcov=True	native	core	certified	—	—
sp.margins_at	—	—	vcov=True	native	core	certified	—	—
sp.contrast	—	—	—	native	core	certified	—	—

Limitations are listed in full by sp.describe_function(name)['limitations'] and in every MCP tool description.