GPC_cpp call for each endpoint and each strata the pairwise comparison function suited to the type of endpoint and store the results.
GPC_cpp(
endpoint,
status,
indexC,
posC,
indexT,
posT,
threshold,
threshold0,
restriction,
weightEndpoint,
weightObs,
method,
pool,
op,
D,
D_UTTE,
n_strata,
nUTTE_analyzedPeron_M1,
index_endpoint,
index_status,
index_UTTE,
list_survTimeC,
list_survTimeT,
list_survJumpC,
list_survJumpT,
list_lastSurv,
p_C,
p_T,
iid_survJumpC,
iid_survJumpT,
zeroPlus,
correctionUninf,
hierarchical,
hprojection,
neutralAsUninf,
addHalfNeutral,
keepScore,
precompute,
paired,
returnIID,
debug
)GPC2_cpp(
endpoint,
status,
indexC,
posC,
indexT,
posT,
threshold,
threshold0,
restriction,
weightEndpoint,
weightObs,
method,
pool,
op,
D,
D_UTTE,
n_strata,
nUTTE_analyzedPeron_M1,
index_endpoint,
index_status,
index_UTTE,
list_survTimeC,
list_survTimeT,
list_survJumpC,
list_survJumpT,
list_lastSurv,
p_C,
p_T,
iid_survJumpC,
iid_survJumpT,
zeroPlus,
correctionUninf,
hierarchical,
hprojection,
neutralAsUninf,
addHalfNeutral,
keepScore,
precompute,
paired,
returnIID,
debug
)
A matrix containing the values of each endpoint (in columns) for each observation (in rows).
A matrix containing the values of the status variables relative to each endpoint (in columns) for each observation (in rows).
A list containing, for each strata, which rows of the endpoint and status matrices corresponds to the control observations. Not unique when bootstraping.
A list containing, for each strata, the unique identifier of each control observations.
A list containing, for each strata, which rows of the endpoint and status matrices corresponds to the treatment observations. Not unique when bootstraping.
A list containing, for each strata, the unique identifier of each treatment observations.
Store the thresholds associated to each endpoint. Must have length D. The threshold is ignored for binary endpoints.
Was the 'original' threshold 0. Must have length D. Ignored for binary endpoints.
Store the restriction time associated to each endpoint. Must have length D.
Store the weight associated to each endpoint. Must have length D.
A vector containing the weight associated to each observation.
The index of the method used to score the pairs. Must have length D. 1 for binary/continuous, 2 for Gaussian, 3/4 for Gehan (left or right-censoring), and 5/6 for Peron (right-censoring survival or competing risks).
The index of the method used to pool results across strata. Can be 0 (weight inversely proportional to the sample size), 1 (Mantel Haenszel weights), 2 (equal weights), 3 (precision weights)
The index of the operator used to score the pairs. Must have length D. 1 for larger is beter, -1 for smaller is better.
The number of endpoints.
The number of distinct time to event endpoints.
The number of strata.
The number of unique time-to-event endpoints that have been analyzed the Peron scoring rule before the current endpoint. Must have length D.
The position of the endpoint at each priority in the argument endpoint. Must have length D.
The position of the status at each priority in the argument status. Must have length D.
The position, among all the unique tte endpoints, of the TTE endpoints. Equals -1 for non tte endpoints. Must have length n_TTE.
A list of matrix containing the survival estimates (-threshold, 0, +threshold ...) for each event of the control group (in rows).
A list of matrix containing the survival estimates (-threshold, 0, +threshold ...) for each event of the treatment group (in rows).
A list of matrix containing the survival estimates and survival jumps when the survival for the control arm jumps.
A list of matrix containing the survival estimates and survival jumps when the survival for the treatment arm jumps.
A list of matrix containing the last survival estimate in each strata (rows) and treatment group (columns).
Number of nuisance parameter in the survival model for the control group, for each endpoint and strata
Number of nuisance parameter in the survival model for the treatment group, for each endpoint and strata
A list of matrix containing the iid of the survival estimates in the control group.
A list of matrix containing the iid of the survival estimates in the treatment group.
Value under which doubles are considered 0?
Should the uninformative weight be re-distributed to favorable and unfavorable?
Should only the uninformative pairs be analyzed at the lower priority endpoints (hierarchical GPC)? Otherwise all pairs will be compaired for all endpoint (full GPC).
Order of the H-projection used to compute the variance.
Should paired classified as neutral be re-analyzed using endpoints of lower priority?
Should half of the neutral score be added to the favorable and unfavorable scores?
Should the result of each pairwise comparison be kept?
Have the integrals relative to the survival be already computed and stored in list_survTimeC/list_survTimeT and list_survJumpC/list_survJumpT (derivatives)
In case of paired data, the variance of the summary statistic across strata will be added to the variance of the pooled statistic.
Should the iid be computed? Second element: is there any nuisance parameter?
Print messages tracing the execution of the function to help debugging. The amount of messages increase with the value of debug (0-5).
Brice Ozenne
GPC_cpp implements GPC looping first over endpoints and then over pairs. To handle multiple endpoints, it stores some of the results which can be memory demanding when considering large sample - especially when computing the iid decomposition. GPC2_cpp implements GPC looping first over pairs and then over endpoints. It has rather minimal memory requirement but does not handle correction for uninformative pairs.