splinetrials_subgroup_analysis objectncs_analysis_subgroup() returns an object of class
splinetrials_subgroup_analysis: a named list with three to seven elements.
These are each tibbles, and they share many of the same
columns and values but are sorted in a different order. Each contains one row
per unique combination of arm, time_scheduled_label, and subgroup found
in the data (see the arguments of ncs_analysis_subgroup()). The values
in columns arm through change_p_value as well as correlation and
optimizer are identical. The two tables' treatment effect analysis results
columns differ in name and content, with between's columns bearing the
prefix diff_subgroup_ and within's columns bearing the prefix diff_arm_
(see the Treatment effects section of ncs_analysis_subgroup()).
Lastly, only within contains the percent slowing analysis results.
between
A 30-column tibble sorted by time, then by arm, then
by subgroup.
Columns:
arm: values of data[[arm]].
time: values of data[[time_scheduled_label]].
subgroup: values of data[[subgroup]].
n: number of times the combination appears in data.
est: mean of data[[response]].
sd: standard deviation of data[[response]].
se: standard error of data[[response]] (i.e., sd / sqrt(n)).
lower: lower bound of confidence interval.
upper: upper bound of confidence interval.
response_est: estimated marginal mean.
response_se: standard error of response_est.
response_df: degrees of freedom used to calculate the confidence
interval for response_est.
response_lower: lower bound of confidence interval for response_est.
response_upper: upper bound of confidence interval for response_est.
change_est: estimated change from baseline.
change_se: standard error of change_est.
change_df: degrees of freedom used for calculating the confidence
interval for and testing the significance of change_est.
change_lower: lower bound of confidence interval for change_est.
change_upper: upper bound of confidence interval for change_est.
change_test_statistic: test statistic measuring the significance of
change_est.
change_p_value: p-value for the significance of change_est.
diff_subgroup_est: treatment effect of subgroup within arm.
diff_subgroup_se: standard error of diff_subgroup_est.
diff_subgroup_df: degrees of freedom used for calculating the confidence
interval for and testing the significance of diff_subgroup_est.
diff_subgroup_lower: lower bound of confidence interval for
diff_subgroup_est.
diff_subgroup_upper: upper bound of confidence interval for
diff_subgroup_est.
diff_subgroup_test_statistic: test statistic measuring the significance
of diff_subgroup_est.
diff_subgroup_p_value: p-value for the significance of
diff_subgroup_est.
correlation: the covariance structure of the analysis model. This is the
same value repeated for each row.
optimizer: invariably mmrm+tmb to indicate that mmrm::mmrm() (which
uses the TMB package) was used to fit the model.
A 33-column tibble sorted by subgroup, then by arm,
then by time.
Columns:
arm: values of data[[arm]].
time: values of data[[time_scheduled_label]].
subgroup: values of data[[subgroup]].
n: number of times the combination appears in data.
est: mean of data[[response]].
sd: standard deviation of data[[response]].
se: standard error of data[[response]] (i.e., sd / sqrt(n)).
lower: lower bound of confidence interval.
upper: upper bound of confidence interval.
response_est: estimated marginal mean.
response_se: standard error of response_est.
response_df: degrees of freedom used for calculating the confidence
interval for response_est.
response_lower: lower bound of confidence interval for response_est.
response_upper: upper bound of confidence interval for response_est.
change_est: estimated change from baseline.
change_se: standard error of change_est.
change_df: degrees of freedom for calculating the confidence interval
for and estimating the significance of change_est.
change_lower: lower bound of confidence interval for change_est.
change_upper: upper bound of confidence interval for change_est.
change_test_statistic: test statistic measuring the significance of
change_est.
change_p_value: p-value for the significance of change_est.
diff_arm_est: treatment effect of arm within subgroup.
diff_arm_se: standard error of diff_arm_est.
diff_arm_df: degrees of freedom for calculating the confidence interval
for and testing the significance of diff_arm_est.
diff_arm_lower: lower bound of confidence interval for diff_arm_est.
diff_arm_upper: upper bound of confidence interval for diff_arm_est.
diff_arm_test_statistic: test statistic measuring the significance of
diff_arm_est.
diff_arm_p_value: p-value for the significance of diff_arm_est.
percent_slowing_est: estimated percent slowing.
percent_slowing_lower: lower bound of confidence interval for
percent_slowing_est.
percent_slowing_upper: upper bound of confidence interval for
percent_slowing_est.
correlation: the covariance structure of the analysis model. This is the
same value repeated for each row.
optimizer: invariably mmrm+tmb to indicate that mmrm::mmrm() (which
uses the TMB package) was used to fit the model.
A tibble with a row for each term in the model (not
counting any intercepts). Contains the following six columns:
effect: the name of the model term.
chisquare_test_statistic: the Chi-squared test statistic measuring the
significance of the model term.
df: the degrees of freedom used for testing the significance of the
model term.
p_value: the p-value for the significance of the model term.
correlation: the covariance structure of the analysis model. This is the
same value repeated for each row.
optimizer: invariably mmrm+tmb to indicate that mmrm::mmrm() (which
uses the TMB package) was used to fit the model.
This element is only present if subgroup_interaction_test = TRUE.
A 2 by 10 data frame with class anova.mmrm. The first row represents the
"reduced" model and the second row represents the "full" model. The columns
are as follows:
model: c("reduced model", "full model"), identifying the model
associated with each row.
aic: the AIC of the model.
bic: the BIC of the model.
loglik: the log likelihood of the model.
-2*log(l): equal to -2 * loglik.
test_statistic: the test statistic used for testing the significance of
the second-order interaction term(s) between the spline time, subgroup, and
arm. This value is the second element of the column; the first element is
always a missing value.
df: the degrees of freedom used for testing the significance of the
second-order interaction term(s) between the spline term, subgroup, and
arm. This value is the second element of the column; the first element is
always a missing value.
p_value: the p-value for the significance of the second-order
interaction term(s) between the spline term, subgroup, and arm. This
value is the second element of the column; the first element is always a
missing value.
correlation: the covariance structure of the analysis model. This is the
same value repeated for each row.
optimizer: invariably mmrm+tmb to indicate that mmrm::mmrm() (which
uses the TMB package) was used to fit the model.
This element is only present if return_models = TRUE.
An mmrm object: the fitted model used to perform analyses
that produced the between, within, and type3 results.
These elements are only present if subgroup_interaction_test = TRUE and
return_models = TRUE.
Both are mmrm objects: the two maximum-likelihood-estimated
models used to perform the subgroup interaction test whose results are in the
interaction element. See the Subgroup interaction test section of
ncs_analysis_subgroup().
The function ncs_analysis_subgroup(), which produces objects of
this class.