*! Fetches native structured component-inference outputs and validates *! their dimensions, numerical diagnostics, and execution receipts. program define fevc__rust_component_fetch, rclass version 18.0 args handle reference model simulations memorylimit posted level /// outprimitive outcovariance outmcse outspectrum outq1raw /// outsummaries outfolds outcv outreceipt outresults outq1results /// outunits deletion expectedunits expectedgramprobes local modelcode = cond("`model'"=="structured_common",1,2) local referencecode = ("`reference'"=="q1") capture noisily fevc__rust_public_call componentresultv5 `handle', reference(`reference') // The owning fevc program releases the prepared generation on any error; // its private abort routine is not visible from an autoloaded helper. if _rc exit _rc tempname primitive covariance mcse spectrum q1_raw summaries folds cv receipt tempname inference_results q1_results units target_status matrix `target_status' = r(target_variance_status) local joint_status = r(joint_status) local variance_fit = r(variance_fit) matrix `units' = (r(unit_schema),r(unit_deletion),r(independent_units), /// r(nuisance_uncertainty_omitted),r(effective_match_count), /// r(largest_match_mass_share),r(largest_match_leverage), /// r(smallest_maker_denominator),r(mover_units),r(stayer_observations)) matrix colnames `units' = schema deletion independent_units nuisance_omitted /// effective_matches largest_mass_share largest_leverage smallest_maker mover_units stayer_observations matrix `primitive' = r(primitive_covariance) matrix `covariance' = r(covariance) matrix `mcse' = r(trace_mcse) matrix `spectrum' = r(spectrum) if "`reference'"=="q1" matrix `q1_raw' = r(q1) matrix `summaries' = r(variance_summaries) matrix `folds' = r(fold_diagnostics) matrix `cv' = r(cv_diagnostics) matrix `receipt' = (r(schema_version),r(variance_source), /// r(reference_distribution),r(probes),r(counter_atoms), /// r(counter_words),r(peak_forecast_bytes),r(psd_cleanup), /// r(raw_eigen_min),r(raw_eigen_max), /// r(point_correction_identity_error),r(maximum_iterations), /// r(maximum_reduced_residual),r(maximum_complete_residual), /// r(full_residual_tolerance),r(median_absolute_log_ratio), /// r(p90_absolute_log_ratio),r(maximum_absolute_log_ratio), /// r(log_variance_correlation),r(critical_simulations), /// r(maximum_remainder_identity_error),r(solver_columns),r(critical_draws), /// r(joint_status),r(variance_fit),r(gram_probes),r(ordering),r(gram_rcond), /// r(gram_inverse_relres),r(variance_fit_relres),r(positivity_floor), /// r(floored_predictions),r(nonpositive_predictions),r(computed_targets), /// `target_status'[1,2],`target_status'[2,2],`target_status'[3,2],`target_status'[4,2]) matrix colnames `receipt' = schema model reference probes /// counter_atoms counter_words peak psd_cleanup raw_eigen_min /// raw_eigen_max point_identity_error max_iterations /// max_reduced_residual max_complete_residual /// full_residual_tolerance median_abs_log_ratio p90_abs_log_ratio /// maximum_abs_log_ratio log_variance_correlation critical_simulations /// maximum_remainder_identity_error solver_columns critical_draws /// joint_status variance_fit gram_probes ordering gram_rcond gram_relres /// fit_relres positivity_floor floored nonpositive computed q0_worker q0_firm q0_covariance q0_total local ok = rowsof(`primitive')==3 & colsof(`primitive')==3 & /// rowsof(`covariance')==4 & colsof(`covariance')==4 & /// rowsof(`mcse')==3 & colsof(`mcse')==3 & /// rowsof(`spectrum')==4 & colsof(`spectrum')==15 & /// rowsof(`summaries')==2 & colsof(`summaries')==12 & /// rowsof(`folds')==10 & colsof(`folds')==15 & /// rowsof(`cv')==70 & colsof(`cv')==7 & /// rowsof(`target_status')==4 & colsof(`target_status')==2 & /// inlist(`joint_status',0,1,2) & /// `variance_fit'==2 & /// `receipt'[1,1]==5 & `receipt'[1,2]==`modelcode' & /// `receipt'[1,3]==`referencecode' & /// `receipt'[1,4]==`simulations' & `receipt'[1,5]>0 & /// `receipt'[1,6]>0 & `receipt'[1,7]>0 & /// ("$VCKSS_MEMORY_ADVISORY"=="1" | `receipt'[1,7]<=`memorylimit') & `receipt'[1,8]>=0 & /// `receipt'[1,11]<=1e-10 & `receipt'[1,12]>=0 & /// `receipt'[1,13]>=0 & `receipt'[1,14]>=0 & /// `receipt'[1,15]>0 & `receipt'[1,14]<=`receipt'[1,15] & /// `receipt'[1,20]==cond("`reference'"=="q1",max(100000,100*`simulations'),0) & /// `receipt'[1,21]>=0 & `receipt'[1,22]>0 & `receipt'[1,23]>=0 & /// `receipt'[1,23]<=4*`receipt'[1,20] & /// inlist(`units'[1,1],1,2) & `units'[1,2]==cond("`deletion'"=="match",1,2) & /// `units'[1,3]==`expectedunits' & `units'[1,3]>0 & /// `units'[1,4]==("`deletion'"=="match") if "`deletion'"=="match" local ok = `ok' & /// `units'[1,5]>=1 & `units'[1,5]<=`expectedunits'*(1+1e-12) & /// `units'[1,6]>0 & `units'[1,6]<=1 & /// `units'[1,7]>=0 & `units'[1,7]<1 & /// `units'[1,8]>0 & `units'[1,8]<=1 & `units'[1,9]>0 & /// `units'[1,9]+`units'[1,10]==`units'[1,3] else local ok = `ok' & `units'[1,5]==0 & `units'[1,6]==0 & /// `units'[1,7]==0 & `units'[1,8]==0 if "`reference'"=="q1" local ok = `ok' & rowsof(`q1_raw')==4 & /// colsof(`q1_raw')==20 if `ok' { foreach matrix_name in primitive covariance mcse summaries folds cv receipt units target_status { forvalues row = 1/`=rowsof(``matrix_name'')' { forvalues column = 1/`=colsof(``matrix_name'')' { local expected_missing = /// (`joint_status'>0 & inlist("`matrix_name'","primitive","covariance")) | /// (`variance_fit'==2 & inlist("`matrix_name'","summaries","folds","cv")) | /// ("`matrix_name'"=="receipt" & `variance_fit'==2 & inrange(`column',16,19)) | /// ("`matrix_name'"=="receipt" & `variance_fit'==1 & inrange(`column',28,31)) if missing(``matrix_name''[`row',`column']) != `expected_missing' local ok = 0 } } } if `variance_fit'==2 local ok = `ok' & `receipt'[1,26]==`expectedgramprobes' & /// `receipt'[1,27]==cond(`units'[1,10]>0,4,cond("`deletion'"=="match",3,2)) & /// `receipt'[1,28]>0 & `receipt'[1,29]>=0 & /// `receipt'[1,30]>=0 & `receipt'[1,31]>0 & `receipt'[1,32]<`expectedunits' & /// `receipt'[1,33]<=`receipt'[1,32] else local ok = `ok' & `receipt'[1,26]==0 & `receipt'[1,27]==0 & /// `receipt'[1,32]==0 & `receipt'[1,33]==0 local q0computed = 0 forvalues row = 1/4 { local targetcode = `target_status'[`row',2] if !inlist(`targetcode',0,1,4,6) local ok = 0 if `targetcode'==0 & `target_status'[`row',1]<=0 local ok = 0 local q0computed = `q0computed'+(`targetcode'==0) } if "`reference'"=="q0" & `receipt'[1,34]!=`q0computed' local ok = 0 if "`reference'"=="q1" { local expected_critical_draws = 0 local q1computed = 0 forvalues row = 1/4 { local targetstatus = `q1_raw'[`row',17] local q1computed = `q1computed'+(`targetstatus'==0) if !inlist(`targetstatus',0,1,2,3,6) local ok = 0 if inlist(`targetstatus',0,3) local expected_critical_draws = /// `expected_critical_draws'+`receipt'[1,20] forvalues column = 1/20 { local optional = `targetstatus'>0 & /// (inrange(`column',9,13) | `column'==18) if !`optional' & missing(`q1_raw'[`row',`column']) local ok = 0 } if `targetstatus'>0 & (!missing(`q1_raw'[`row',11]) | /// !missing(`q1_raw'[`row',12])) local ok = 0 if reldif(`q1_raw'[`row',1],`posted'[1,`row'])>1e-10 local ok = 0 if abs(`q1_raw'[`row',19]-`q1_raw'[`row',20]-`q1_raw'[`row',7])> /// 1e-10*max(1,abs(`q1_raw'[`row',7])) local ok = 0 if `targetstatus'==0 { if `q1_raw'[`row',3]<=0 | `q1_raw'[`row',7]<=0 | /// `q1_raw'[`row',18]<=0 | `q1_raw'[`row',18]>1 | /// `q1_raw'[`row',10]<=0 | `q1_raw'[`row',11]>`q1_raw'[`row',12] local ok = 0 local determinant = 1-(`q1_raw'[`row',6]/sqrt(`q1_raw'[`row',3])/ /// sqrt(`q1_raw'[`row',7]))^2 if abs(`determinant'-`q1_raw'[`row',18])>1e-10 local ok = 0 local curvature = 2*abs(`spectrum'[`row',1])*`q1_raw'[`row',3]/ /// sqrt(`q1_raw'[`row',7]*`q1_raw'[`row',18]) if reldif(`curvature',`q1_raw'[`row',9])>1e-9 local ok = 0 } } if `receipt'[1,23]!=`expected_critical_draws' local ok = 0 if `receipt'[1,34]!=`q1computed' local ok = 0 } if `receipt'[1,22]!=floor(`receipt'[1,22]) local ok = 0 forvalues row = 1/4 { if `joint_status'==0 & reldif(`target_status'[`row',1],`covariance'[`row',`row'])>1e-12 local ok = 0 if `spectrum'[`row',13]!=128 | `spectrum'[`row',14]!= /// cond("`deletion'"=="observation" | "`reference'"=="q0",512,128) local ok = 0 local spectral_missing_allowed = (`target_status'[`row',2]==6) if "`reference'"=="q1" local spectral_missing_allowed = (`q1_raw'[`row',17]==6) forvalues column = 1/15 { if missing(`spectrum'[`row',`column']) & /// (!`spectral_missing_allowed' | `column'>12) local ok = 0 } } } if `ok' & `joint_status'==0 { local scale = max(1,abs(`covariance'[4,4])) forvalues row = 1/4 { forvalues column = 1/4 { if abs(`covariance'[`row',`column']- /// `covariance'[`column',`row'])>1e-12*`scale' local ok = 0 } } forvalues row = 1/3 { forvalues column = 1/3 { if reldif(`primitive'[`row',`column'],`covariance'[`row',`column'])>1e-12 local ok = 0 } local mapped = `primitive'[`row',1]+`primitive'[`row',2]+ /// 2*`primitive'[`row',3] if abs(`covariance'[`row',4]-`mapped')>1e-12*`scale' local ok = 0 } local mapped44 = `primitive'[1,1]+`primitive'[2,2]+4*`primitive'[3,3]+ /// 2*`primitive'[1,2]+4*`primitive'[1,3]+4*`primitive'[2,3] if abs(`covariance'[4,4]-`mapped44')>1e-12*`scale' local ok = 0 } local names worker_variance firm_variance worker_firm_covariance total_variance matrix rownames `primitive' = worker_variance firm_variance worker_firm_covariance matrix colnames `primitive' = worker_variance firm_variance worker_firm_covariance matrix rownames `covariance' = `names' matrix colnames `covariance' = `names' matrix rownames `mcse' = worker_variance firm_variance worker_firm_covariance matrix colnames `mcse' = worker_variance firm_variance worker_firm_covariance matrix rownames `spectrum' = `names' matrix colnames `spectrum' = lambda1 lambda2 trace2_raw trace2 trace2_mcse /// trace_reconciliation leading_share leading_share_mcse /// remainder_leading_share max_mode_weight_sq leading_residual /// second_residual probes iterations max_influence_share matrix rownames `summaries' = structured_common structured_leverage matrix colnames `summaries' = model observations minimum median maximum /// floor_count floor_share boundary_count boundary_share /// maximum_boundary_excess max_prediction_leverage minimum_fitted_rcond matrix rownames `folds' = common_fold0 common_fold1 common_fold2 common_fold3 /// common_fold4 leverage_fold0 leverage_fold1 leverage_fold2 leverage_fold3 /// leverage_fold4 matrix colnames `folds' = model outer_fold training_observations /// validation_observations active_terms selected_lambda selected_cv_mse /// variance_scale positivity_floor floored_predictions boundary_predictions /// max_boundary_excess max_prediction_leverage fitted_rcond fitted_relres matrix colnames `cv' = model outer_fold grid_index lambda /// validation_observations mse available matrix `inference_results' = J(4,4,.) local zcrit = invnormal(1-(100-`level')/200) forvalues row = 1/4 { local estimate = `posted'[1,`row'] local se = cond(`target_status'[`row',2]==0,sqrt(`target_status'[`row',1]),.) matrix `inference_results'[`row',1] = `estimate' matrix `inference_results'[`row',2] = `se' matrix `inference_results'[`row',3] = `estimate'-`zcrit'*`se' matrix `inference_results'[`row',4] = `estimate'+`zcrit'*`se' } matrix rownames `inference_results' = `names' matrix colnames `inference_results' = estimate se lb ub if "`reference'"=="q1" { matrix rownames `q1_raw' = `names' matrix colnames `q1_raw' = estimate leading_score leading_variance /// leading_component remainder_estimate leading_remainder_cov /// remainder_variance remainder_trace_mcse curvature critical_value /// am_lb am_ub leading_F remainder_influence_share /// recenter_var_b1 remainder_identity_error target_status /// standardized_determinant remainder_influence_variance remainder_trace_variance matrix `q1_results' = J(4,17,.) forvalues row = 1/4 { matrix `q1_results'[`row',1] = `posted'[1,`row'] matrix `q1_results'[`row',2] = `inference_results'[`row',2] matrix `q1_results'[`row',3] = `inference_results'[`row',3] matrix `q1_results'[`row',4] = `inference_results'[`row',4] matrix `q1_results'[`row',5] = `q1_raw'[`row',11] matrix `q1_results'[`row',6] = `q1_raw'[`row',12] matrix `q1_results'[`row',7] = `spectrum'[`row',1] matrix `q1_results'[`row',8] = `spectrum'[`row',7] matrix `q1_results'[`row',9] = `spectrum'[`row',10] matrix `q1_results'[`row',10] = `q1_raw'[`row',3] matrix `q1_results'[`row',11] = `q1_raw'[`row',6] matrix `q1_results'[`row',12] = `q1_raw'[`row',7] matrix `q1_results'[`row',13] = `q1_raw'[`row',2] matrix `q1_results'[`row',14] = `q1_raw'[`row',5] matrix `q1_results'[`row',15] = `q1_raw'[`row',13] matrix `q1_results'[`row',16] = `q1_raw'[`row',9] matrix `q1_results'[`row',17] = `q1_raw'[`row',10] } matrix rownames `q1_results' = `names' matrix colnames `q1_results' = estimate highrank_se wald_lb wald_ub /// am_lb am_ub lambda1 eigen_share max_weight_sq var_b1 /// cov_b1_theta1 var_theta1 b1 theta1 F curvature critical_value } if !`ok' { exit 498 } matrix `outprimitive' = `primitive' matrix `outcovariance' = `covariance' matrix `outmcse' = `mcse' matrix `outspectrum' = `spectrum' matrix `outsummaries' = `summaries' matrix `outfolds' = `folds' matrix `outcv' = `cv' matrix `outreceipt' = `receipt' matrix `outunits' = `units' matrix `outresults' = `inference_results' if "`reference'"=="q1" { matrix `outq1raw' = `q1_raw' matrix `outq1results' = `q1_results' } return scalar peak = `receipt'[1,7] end