Introduction to rsimsum

Alessandro Gasparini

2018-04-04

rsimsum

rsimsum is an R package that can compute summary statistics from simulation studies. It is a port to R of the user-written command simsum in Stata (White I.R., 2010).

The aim of rsimsum is helping reporting of simulation studies, including understanding the role of chance in results of simulation studies. Specifically, rsimsum can compute Monte Carlo standard errors of summary statistics, defined as the standard deviation of the estimated summary statistic.

Formula for summary statistics and Monte Carlo standard errors are presented in the next section.

Notation

We will use th following notation throughout this vignette:

Summary statistics

The first summary statistic of interest is bias, which quantifies whether the estimator targets the true value \(\theta\) on average. Bias is calculated as:

\[\text{Bias} = \frac{1}{n_{\text{sim}}} \sum_{i = 1} ^ {n_{\text{sim}}} \hat{\theta}_i - \theta\]

The Monte Carlo standard error of bias is calculated as:

\[\text{MCSE(Bias)} = \sqrt{\frac{\frac{1}{n_{\text{sim}} - 1} \sum_{i = 1} ^ {n_{\text{sim}}} (\hat{\theta}_i - \bar{\theta}) ^ 2}{n_{\text{sim}}}}\]

The empirical standard error of \(\theta\) depends only on \(\hat{\theta}\) and does not require any knowledge of \(\theta\). It estimates the standard deviation of \(\hat{\theta}\) over the \(n_{\text{sim}}\) replications:

\[\text{Empirical SE} = \sqrt{\frac{1}{n_{\text{sim}} - 1} \sum_{i = 1} ^ {n_{\text{sim}}} (\hat{\theta}_i - \bar{\theta}) ^ 2}\]

The Monte Carlo standard error is calculated as:

\[\text{MCSE(Emp. SE)} = \frac{\text{Emp. SE}}{\sqrt{2 (n_{\text{sim}} - 1)}}\]

When comparing different methods, the relative precision of a given method B against a reference method A is computed as:

\[\text{Relative % increase in precision} = \left( \frac{\text{Emp. SE}_A}{\text{Emp. SE}_B} \right) ^ 2\]

Its (approximated) Monte Carlo standard error is:

\[\text{MCSE(Relative % increase in precision)} \simeq 2 \frac{\text{Var}(\hat{\theta}_A)}{\text{Var}(\hat{\theta}_B)} \sqrt{\frac{1 - \rho^2_{AB}}{n_{\text{sim}} - 1}}\]

\(\rho^2_{AB}\) is the correlation of \(\hat{\theta}_A\) and \(\hat{\theta}_B\).

A measure that takes into account both precision and accuracy of a method is the mean squared error, which is the sum of the squared bias and variance of \(\hat{\theta}\):

\[\text{MSE} = \frac{1}{n_{\text{sim}}} = \sum_{i = 1} ^ {n_{\text{sim}}} (\hat{\theta}_i - \theta) ^ 2\]

The Monte Carlo standard error is:

\[\text{MCSE(MSE)} = \sqrt{\frac{\sum_{i = 1} ^ {n_{\text{sim}}} \left[ (\hat{\theta}_i - \theta) ^2 - \text{MSE} \right] ^ 2}{n_{\text{sim}} (n_{\text{sim}} - 1)}}\]

The model based standard error is computed by averaging the estimated standard errors for each replication:

\[\text{Model SE} = \sqrt{\frac{1}{n_{\text{sim}}} \sum_{i = 1} ^ {n_{\text{sim}}} \widehat{\text{Var}}(\hat{\theta}_i))}\]

Its (approximated) Monte Carlo standard error is computed as:

\[\text{MCSE(Model SE)} \simeq \sqrt{\frac{\text{Var}[\widehat{\text{Var}}(\hat{\theta}_i)]}{4 n_{\text{sim}} \widehat{\text{Model SE}}}}\]

The model standard error targets the empirical standard error. Hence, the relative error in the model standard error is an informative performance measure:

\[\text{Relative % error in model SE} = 100 \left( \frac{\text{Model SE}}{\text{Empirical SE}} - 1\right)\]

Its Monte Carlo standard error is computed as:

\[\text{MCSE(Relative % error in model SE)} = 100 \left( \frac{\text{Model SE}}{\text{Empirical SE}} \right) \sqrt{\frac{\text{Var}[\widehat{\text{Var}}(\hat{\theta}_i)]}{4 n_{\text{sim}} \widehat{\text{Model SE}} ^ 4} + \frac{1}{2(n_{\text{sim}} - 1)}}\]

Coverage is another key property of an estimator. It is defined as the probability that a confidence interval contains the true value \(\theta\), and computed as:

\[\text{Coverage} = \frac{1}{n_{\text{sim}}} \sum_{i = 1} ^ {n_{\text{sim}}} I(\hat{\theta}_{i, \text{low}} \le \theta \le \hat{\theta}_{i, \text{upp}})\]

where \(I(\cdot)\) is the indicator function. The Monte Carlo standard error is computed as:

\[\text{MCSE(Coverage)} = \sqrt{\frac{\text{Coverage} \times (1 - \text{Coverage})}{n_{\text{sim}}}}\]

Under coverage is to be expected if:

  1. \(\text{Bias} \ne 0\), or
  2. \(\text{Models SE} < \text{Empirical SE}\), or
  3. the distribution of \(\hat{\theta}\) is not normal and intervals have been constructed assuming normality, or
  4. \(\widehat{\text{Var}}(\hat{\theta}_i)\) is too variable

Over coverage occurs as a result of \(\text{Models SE} > \text{Empirical SE}\).

As under coverage may be a result of bias, another useful summary statistic is bias corrected coverage:

\[\text{Bias corrected coverage} = \frac{1}{n_{\text{sim}}} \sum_{i = 1} ^ {n_{\text{sim}}} I(\hat{\theta}_{i, \text{low}} \le \bar{\theta} \le \hat{\theta}_{i, \text{upp}}) \]

The Monte Carlo standard error is analogously as coverage:

\[\text{MCSE(Bias corrected coverage)} = \sqrt{\frac{\text{Bias corrected coverage} \times (1 - \text{Bias corrected coverage})}{n_{\text{sim}}}}\]

Finally, power of a significance test at the \(\alpha\) level is defined as:

\[\text{Power} = \frac{1}{n_{\text{sim}}} \sum_{i = 1} ^ {n_{\text{sim}}} I \left[ |\hat{\theta}_i| \ge z_{\alpha/2} * \sqrt{\widehat{\text{Var}}(\hat{\theta_i})} \right]\]

The Monte Carlo standard error is analogously as coverage:

\[\text{MCSE(Power)} = \sqrt{\frac{\text{Power} \times (1 - \text{Power})}{n_{\text{sim}}}}\]

Further information on summary statistics for simulation studies can be found in White (2010) and Morris, White, and Crowther (2017).

Example 1: Simulation study on missing data

With this example dataset included in rsimsum we aim to summarise a simulation study comparing different ways to handle missing covariates when fitting a Cox model (White and Royston, 2009). One thousand datasets were simulated, each containing normally distributed covariates \(x\) and \(z\) and time-to-event outcome. Both covariates has \(20\%\) of their values deleted independently of all other variables so the data became missing completely at random (Little and Rubin, 2002). Each simulated dataset was analysed in three ways. A Cox model was fit to the complete cases (CC). Then two methods of multiple imputation using chained equations (van Buuren, Boshuizen, and Knook, 1999) were used. The MI_LOGT method multiply imputes the missing values of \(x\) and \(z\) with the outcome included as \(\log(t)\) and \(d\), where \(t\) is the survival time and \(d\) is the event indicator. The MI_T method is the same except that \(\log(t)\) is replaced by \(t\) in the imputation model.

We load the data in the usual way:

library(rsimsum)
#> 
#> Attaching package: 'rsimsum'
#> The following object is masked from 'package:utils':
#> 
#>     zip
data("MIsim", package = "rsimsum")

Let’s have a look at the first 10 rows of the dataset:

head(MIsim, n = 10)
#>    dataset  method         b        se
#> 1        1      CC 0.7067682 0.1465100
#> 2        1    MI_T 0.6841882 0.1255043
#> 3        1 MI_LOGT 0.7124795 0.1410814
#> 4        2      CC 0.3485008 0.1599879
#> 5        2    MI_T 0.4060082 0.1409831
#> 6        2 MI_LOGT 0.4287003 0.1358589
#> 7        3      CC 0.6495075 0.1521568
#> 8        3    MI_T 0.5028701 0.1300780
#> 9        3 MI_LOGT 0.5604051 0.1168512
#> 10       4      CC 0.4320534 0.1262853

The included variables are:

str(MIsim)
#> Classes 'tbl_df', 'tbl' and 'data.frame':    3000 obs. of  4 variables:
#>  $ dataset: num  1 1 1 2 2 2 3 3 3 4 ...
#>  $ method : chr  "CC" "MI_T" "MI_LOGT" "CC" ...
#>  $ b      : num  0.707 0.684 0.712 0.349 0.406 ...
#>  $ se     : num  0.147 0.126 0.141 0.16 0.141 ...
#>  - attr(*, "label")= chr "simsum example: data from a simulation study comparing 3 ways to handle missing"

We summarise the results of the simulation study by method using the simsum function:

s1 <- simsum(data = MIsim, estvarname = "b", true = 0.50, se = "se", methodvar = "method", ref = "CC")

Using the default settings, Monte Carlo standard errors are computed and returned.

ss1 <- summary(s1)
ss1
#> 
#> Call:
#>  simsum(data = MIsim, estvarname = "b", true = 0.5, se = "se", 
#>     methodvar = "method", ref = "CC")
#> 
#> Method variable: method 
#>  Unique methods: CC, MI_LOGT, MI_T 
#>  Reference method: CC 
#> By factors: none
#> 
#> Summary statistics:
#> 
#>  Method = CC 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate    0.5168     NA        NA        NA
#>                       Median point estimate    0.5070     NA        NA        NA
#>                      Average standard error    0.0216     NA        NA        NA
#>                       Median standard error    0.0211     NA        NA        NA
#>                      Bias in point estimate    0.0168 0.0048    0.0074    0.0261
#>                    Empirical standard error    0.1511 0.0034    0.1445    0.1577
#>                          Mean squared error    0.0231 0.0011    0.0209    0.0253
#>   % gain in precision relative to method CC    1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error    0.1471 0.0005    0.1461    0.1481
#>          Relative % error in standard error   -2.6594 2.2049   -6.9810    1.6622
#>                  Coverage of nominal 95% CI    0.9430 0.0073    0.9286    0.9574
#>   Bias corrected coverage of nominal 95% CI    0.9400 0.0075    0.9253    0.9547
#>                      Power of 5% level test    0.9460 0.0071    0.9320    0.9600
#> 
#>  Method = MI_LOGT 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate    0.5009     NA        NA        NA
#>                       Median point estimate    0.4969     NA        NA        NA
#>                      Average standard error    0.0182     NA        NA        NA
#>                       Median standard error    0.0172     NA        NA        NA
#>                      Bias in point estimate    0.0009 0.0042   -0.0073    0.0091
#>                    Empirical standard error    0.1320 0.0030    0.1262    0.1378
#>                          Mean squared error    0.0174 0.0009    0.0157    0.0191
#>   % gain in precision relative to method CC    1.3105 0.0394    1.2333    1.3876
#>                  Model-based standard error    0.1349 0.0006    0.1338    0.1361
#>          Relative % error in standard error    2.2233 2.3318   -2.3469    6.7935
#>                  Coverage of nominal 95% CI    0.9490 0.0070    0.9354    0.9626
#>   Bias corrected coverage of nominal 95% CI    0.9490 0.0070    0.9354    0.9626
#>                      Power of 5% level test    0.9690 0.0055    0.9583    0.9797
#> 
#>  Method = MI_T 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate    0.4988     NA        NA        NA
#>                       Median point estimate    0.4939     NA        NA        NA
#>                      Average standard error    0.0179     NA        NA        NA
#>                       Median standard error    0.0169     NA        NA        NA
#>                      Bias in point estimate   -0.0012 0.0043   -0.0095    0.0071
#>                    Empirical standard error    0.1344 0.0030    0.1285    0.1403
#>                          Mean squared error    0.0181 0.0009    0.0163    0.0198
#>   % gain in precision relative to method CC    1.2637 0.0384    1.1884    1.3390
#>                  Model-based standard error    0.1338 0.0006    0.1327    0.1350
#>          Relative % error in standard error   -0.4412 2.2690   -4.8883    4.0059
#>                  Coverage of nominal 95% CI    0.9430 0.0073    0.9286    0.9574
#>   Bias corrected coverage of nominal 95% CI    0.9430 0.0073    0.9286    0.9574
#>                      Power of 5% level test    0.9630 0.0060    0.9513    0.9747

Tabulating summary statistics

It is straightforward to produce a table of summary statistics for use in an R Markdown document:

library(knitr)
kable(get_data(ss1))
stat est mcse method lower upper
nsim 1000.0000000 NA CC NA NA
thetamean 0.5167662 NA CC NA NA
thetamedian 0.5069935 NA CC NA NA
se2mean 0.0216373 NA CC NA NA
se2median 0.0211425 NA CC NA NA
bias 0.0167662 0.0047787 CC 0.0074001 0.0261322
empse 0.1511150 0.0033807 CC 0.1444889 0.1577411
mse 0.0230940 0.0011338 CC 0.0208717 0.0253163
relprec 1.0000000 0.0000000 CC 1.0000000 1.0000000
modelse 0.1470963 0.0005274 CC 0.1460626 0.1481300
relerror -2.6593842 2.2049438 CC -6.9809947 1.6622263
cover 0.9430000 0.0073315 CC 0.9286305 0.9573695
bccover 0.9400000 0.0075100 CC 0.9252807 0.9547193
power 0.9460000 0.0071473 CC 0.9319915 0.9600085
nsim 1000.0000000 NA MI_LOGT NA NA
thetamean 0.5009231 NA MI_LOGT NA NA
thetamedian 0.4969223 NA MI_LOGT NA NA
se2mean 0.0182091 NA MI_LOGT NA NA
se2median 0.0172157 NA MI_LOGT NA NA
bias 0.0009231 0.0041744 MI_LOGT -0.0072586 0.0091048
empse 0.1320064 0.0029532 MI_LOGT 0.1262182 0.1377947
mse 0.0174091 0.0008813 MI_LOGT 0.0156818 0.0191364
relprec 1.3104634 0.0393747 MI_LOGT 1.2332904 1.3876365
modelse 0.1349413 0.0006046 MI_LOGT 0.1337563 0.1361263
relerror 2.2232593 2.3317773 MI_LOGT -2.3469401 6.7934588
cover 0.9490000 0.0069569 MI_LOGT 0.9353647 0.9626353
bccover 0.9490000 0.0069569 MI_LOGT 0.9353647 0.9626353
power 0.9690000 0.0054808 MI_LOGT 0.9582579 0.9797421
nsim 1000.0000000 NA MI_T NA NA
thetamean 0.4988092 NA MI_T NA NA
thetamedian 0.4939111 NA MI_T NA NA
se2mean 0.0179117 NA MI_T NA NA
se2median 0.0169319 NA MI_T NA NA
bias -0.0011908 0.0042510 MI_T -0.0095226 0.0071409
empse 0.1344277 0.0030074 MI_T 0.1285333 0.1403221
mse 0.0180542 0.0009112 MI_T 0.0162682 0.0198401
relprec 1.2636816 0.0384238 MI_T 1.1883724 1.3389909
modelse 0.1338346 0.0005856 MI_T 0.1326867 0.1349824
relerror -0.4412233 2.2689748 MI_T -4.8883321 4.0058856
cover 0.9430000 0.0073315 MI_T 0.9286305 0.9573695
bccover 0.9430000 0.0073315 MI_T 0.9286305 0.9573695
power 0.9630000 0.0059692 MI_T 0.9513006 0.9746994

Using get_data() in combination with R packages such as xtable, kableExtra, tables can yield a variety of tables that should suit most purposes.

More information on producing tables directly from R can be found in the CRAN Task View on Reproducible Research.

Plotting summary statistics

In this section, we show how to plot and compare summary statistics using the popular R package ggplot.

Plotting bias by method with \(95\%\) confidence intervals based on Monte Carlo standard errors:

library(ggplot2)
ggplot(get_data(ss1, sstat = "bias"), aes(x = method, y = est, ymin = lower, ymax = upper)) +
  geom_hline(yintercept = 0, color = "red", lty = "dashed") +
  geom_point() +
  geom_errorbar(width = 1 / 3) +
  theme_bw() +
  labs(x = "Method", y = "Bias")

Conversely, say we want to visually compare coverage for the three methods compared with this simulation study:

ggplot(get_data(ss1, sstat = "cover"), aes(x = method, y = est, ymin = lower, ymax = upper)) +
  geom_hline(yintercept = 0.95, color = "red", lty = "dashed") +
  geom_point() +
  geom_errorbar(width = 1 / 3) +
  coord_cartesian(ylim = c(0, 1)) +
  theme_bw() +
  labs(x = "Method", y = "Coverage")

Dropping large estimates and standard errors

rsimsum allows to automatically drop estimates and standard errors that are larger than a predefined value. Specifically, the argument of simsum that control this behaviour are dropbig, max and semax.

Set dropbig to TRUE, and then standardised estimates larger than max in absolute value will be dropped; standard errors larger than semax times the average standard error will be dropped too.

For instance, say we want to drop standardised estimates larger than \(4\) in absolute value and standard errors larger than \(1.5\) times the average standard error:

s1.2 <- simsum(data = MIsim, estvarname = "b", true = 0.50, se = "se", methodvar = "method", ref = "CC", dropbig = TRUE, max = 4, semax = 1.5)

Were any estimates dropped?

dropbig(s1.2)
#> Dropped point estimates:
#> # A tibble: 2 x 4
#>   dataset method       b    se
#>     <dbl> <chr>    <dbl> <dbl>
#> 1    761. MI_T    -0.148 0.143
#> 2    761. MI_LOGT -0.108 0.159
#> 
#> Dropped standard errors:
#> # A tibble: 8 x 4
#>   dataset method      b    se
#>     <dbl> <chr>   <dbl> <dbl>
#> 1     19. MI_T    0.460 0.216
#> 2    159. MI_T    0.860 0.228
#> 3    211. MI_T    0.464 0.215
#> 4    256. CC      0.278 0.219
#> 5    417. CC      0.601 0.210
#> 6    417. MI_LOGT 0.653 0.223
#> 7    551. MI_LOGT 0.543 0.227
#> 8    585. MI_LOGT 0.737 0.207

Yes they were!

Everything else works analogously as before; for instance, to summarise the results:

summary(s1.2)
#> 
#> Call:
#>  simsum(data = MIsim, estvarname = "b", true = 0.5, se = "se", 
#>     methodvar = "method", ref = "CC", dropbig = TRUE, max = 4, 
#>     semax = 1.5)
#> 
#> Method variable: method 
#>  Unique methods: CC, MI_LOGT, MI_T 
#>  Reference method: CC 
#> By factors: none
#> 
#> Summary statistics:
#> 
#>  Method = CC 
#>                                             Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 998.0000     NA        NA        NA
#>                      Average point estimate   0.5169     NA        NA        NA
#>                       Median point estimate   0.5070     NA        NA        NA
#>                      Average standard error   0.0216     NA        NA        NA
#>                       Median standard error   0.0211     NA        NA        NA
#>                      Bias in point estimate   0.0169 0.0048    0.0076    0.0263
#>                    Empirical standard error   0.1511 0.0034    0.1444    0.1577
#>                          Mean squared error   0.0231 0.0011    0.0209    0.0253
#>   % gain in precision relative to method CC   1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error   0.1469 0.0005    0.1459    0.1479
#>          Relative % error in standard error  -2.7304 2.2043   -7.0508    1.5900
#>                  Coverage of nominal 95% CI   0.9429 0.0073    0.9285    0.9573
#>   Bias corrected coverage of nominal 95% CI   0.9399 0.0075    0.9251    0.9546
#>                      Power of 5% level test   0.9469 0.0071    0.9330    0.9608
#> 
#>  Method = MI_LOGT 
#>                                             Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 996.0000     NA        NA        NA
#>                      Average point estimate   0.5011     NA        NA        NA
#>                       Median point estimate   0.4960     NA        NA        NA
#>                      Average standard error   0.0181     NA        NA        NA
#>                       Median standard error   0.0172     NA        NA        NA
#>                      Bias in point estimate   0.0011 0.0041   -0.0070    0.0092
#>                    Empirical standard error   0.1305 0.0029    0.1248    0.1363
#>                          Mean squared error   0.0170 0.0008    0.0154    0.0186
#>   % gain in precision relative to method CC   1.3389 0.0404    1.2596    1.4181
#>                  Model-based standard error   0.1346 0.0006    0.1335    0.1357
#>          Relative % error in standard error   3.0930 2.3522   -1.5173    7.7033
#>                  Coverage of nominal 95% CI   0.9498 0.0069    0.9362    0.9634
#>   Bias corrected coverage of nominal 95% CI   0.9498 0.0069    0.9362    0.9634
#>                      Power of 5% level test   0.9699 0.0054    0.9593    0.9805
#> 
#>  Method = MI_T 
#>                                             Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 996.0000     NA        NA        NA
#>                      Average point estimate   0.4992     NA        NA        NA
#>                       Median point estimate   0.4941     NA        NA        NA
#>                      Average standard error   0.0178     NA        NA        NA
#>                       Median standard error   0.0169     NA        NA        NA
#>                      Bias in point estimate  -0.0008 0.0042   -0.0091    0.0074
#>                    Empirical standard error   0.1326 0.0030    0.1268    0.1384
#>                          Mean squared error   0.0176 0.0008    0.0160    0.0192
#>   % gain in precision relative to method CC   1.2973 0.0398    1.2193    1.3752
#>                  Model-based standard error   0.1335 0.0006    0.1324    0.1346
#>          Relative % error in standard error   0.6505 2.2942   -3.8460    5.1471
#>                  Coverage of nominal 95% CI   0.9438 0.0073    0.9295    0.9581
#>   Bias corrected coverage of nominal 95% CI   0.9438 0.0073    0.9295    0.9581
#>                      Power of 5% level test   0.9639 0.0059    0.9523    0.9754

Example 2: Simulation study on survival modelling

data("relhaz", package = "rsimsum")

Let’s have a look at the first 10 rows of the dataset:

head(relhaz, n = 10)
#>    dataset  n    baseline       theta        se model
#> 1        1 50 Exponential -0.88006151 0.3330172   Cox
#> 2        2 50 Exponential -0.81460242 0.3253010   Cox
#> 3        3 50 Exponential -0.14262887 0.3050516   Cox
#> 4        4 50 Exponential -0.33251820 0.3144033   Cox
#> 5        5 50 Exponential -0.48269940 0.3064726   Cox
#> 6        6 50 Exponential -0.03160756 0.3097203   Cox
#> 7        7 50 Exponential -0.23578090 0.3121350   Cox
#> 8        8 50 Exponential -0.05046332 0.3136058   Cox
#> 9        9 50 Exponential -0.22378715 0.3066037   Cox
#> 10      10 50 Exponential -0.45326446 0.3330173   Cox

The included variables are:

str(relhaz)
#> 'data.frame':    12000 obs. of  6 variables:
#>  $ dataset : int  1 2 3 4 5 6 7 8 9 10 ...
#>  $ n       : num  50 50 50 50 50 50 50 50 50 50 ...
#>  $ baseline: chr  "Exponential" "Exponential" "Exponential" "Exponential" ...
#>  $ theta   : num  -0.88 -0.815 -0.143 -0.333 -0.483 ...
#>  $ se      : num  0.333 0.325 0.305 0.314 0.306 ...
#>  $ model   : chr  "Cox" "Cox" "Cox" "Cox" ...

rsimsum can summarise results from simulation studies with plenty of data-generating mechanisms. For instance, with this example we show how to compute summary statistics by baseline hazard function and sample size.

In order to summarise results by data-generating factors, it is sufficient to define the “by” factors in the call to simsum:

s2 <- simsum(data = relhaz, estvarname = "theta", true = -0.50, se = "se", methodvar = "model", by = c("baseline", "n"))
#> `ref` was not specified, Cox set as the reference
s2
#> 
#> Call:
#>  simsum(data = relhaz, estvarname = "theta", true = -0.5, se = "se", 
#>     methodvar = "model", by = c("baseline", "n"))
#> 
#> Method variable: model 
#>  Unique methods: Cox, Exp, RP(2) 
#>  Reference method: Cox 
#> 
#> By factors: baseline, n 
#> 
#> Monte Carlo standard errors were computed.

Summarising the results will be printed out for each method and combination of data-generating factors:

ss2 <- summary(s2)
ss2
#> 
#> Call:
#>  simsum(data = relhaz, estvarname = "theta", true = -0.5, se = "se", 
#>     methodvar = "model", by = c("baseline", "n"))
#> 
#> Method variable: model 
#>  Unique methods: Cox, Exp, RP(2) 
#>  Reference method: Cox 
#> By factors: baseline, n 
#> 
#> Summary statistics:
#> 
#>  Method = Cox, baseline = Exponential, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5004     NA        NA        NA
#>                       Median point estimate   -0.4969     NA        NA        NA
#>                      Average standard error    0.0194     NA        NA        NA
#>                       Median standard error    0.0194     NA        NA        NA
#>                      Bias in point estimate   -0.0004 0.0043   -0.0088    0.0081
#>                    Empirical standard error    0.1363 0.0030    0.1303    0.1422
#>                          Mean squared error    0.0185 0.0009    0.0168    0.0203
#>  % gain in precision relative to method Cox    1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error    0.1394 0.0001    0.1393    0.1396
#>          Relative % error in standard error    2.3293 2.2894   -2.1578    6.8164
#>                  Coverage of nominal 95% CI    0.9510 0.0068    0.9376    0.9644
#>   Bias corrected coverage of nominal 95% CI    0.9520 0.0068    0.9388    0.9652
#>                      Power of 5% level test    0.9570 0.0064    0.9444    0.9696
#> 
#>  Method = Exp, baseline = Exponential, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.4996     NA        NA        NA
#>                       Median point estimate   -0.5004     NA        NA        NA
#>                      Average standard error    0.0191     NA        NA        NA
#>                       Median standard error    0.0190     NA        NA        NA
#>                      Bias in point estimate    0.0004 0.0043   -0.0080    0.0088
#>                    Empirical standard error    0.1350 0.0030    0.1291    0.1409
#>                          Mean squared error    0.0182 0.0009    0.0165    0.0199
#>  % gain in precision relative to method Cox    1.0192 0.0093    1.0009    1.0375
#>                  Model-based standard error    0.1381 0.0001    0.1379    0.1382
#>          Relative % error in standard error    2.2813 2.2882   -2.2035    6.7661
#>                  Coverage of nominal 95% CI    0.9540 0.0066    0.9410    0.9670
#>   Bias corrected coverage of nominal 95% CI    0.9540 0.0066    0.9410    0.9670
#>                      Power of 5% level test    0.9570 0.0064    0.9444    0.9696
#> 
#>  Method = RP(2), baseline = Exponential, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5015     NA        NA        NA
#>                       Median point estimate   -0.5000     NA        NA        NA
#>                      Average standard error    0.0194     NA        NA        NA
#>                       Median standard error    0.0193     NA        NA        NA
#>                      Bias in point estimate   -0.0015 0.0043   -0.0100    0.0069
#>                    Empirical standard error    0.1363 0.0030    0.1303    0.1423
#>                          Mean squared error    0.0186 0.0009    0.0168    0.0203
#>  % gain in precision relative to method Cox    0.9996 0.0034    0.9929    1.0063
#>                  Model-based standard error    0.1392 0.0001    0.1391    0.1394
#>          Relative % error in standard error    2.1570 2.2855   -2.3225    6.6365
#>                  Coverage of nominal 95% CI    0.9520 0.0068    0.9388    0.9652
#>   Bias corrected coverage of nominal 95% CI    0.9520 0.0068    0.9388    0.9652
#>                      Power of 5% level test    0.9590 0.0063    0.9467    0.9713
#> 
#>  Method = Cox, baseline = Weibull, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5057     NA        NA        NA
#>                       Median point estimate   -0.5031     NA        NA        NA
#>                      Average standard error    0.0173     NA        NA        NA
#>                       Median standard error    0.0172     NA        NA        NA
#>                      Bias in point estimate   -0.0057 0.0040   -0.0136    0.0022
#>                    Empirical standard error    0.1274 0.0029    0.1219    0.1330
#>                          Mean squared error    0.0163 0.0008    0.0147    0.0178
#>  % gain in precision relative to method Cox    1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error    0.1316 0.0001    0.1315    0.1317
#>          Relative % error in standard error    3.2717 2.3104   -1.2566    7.8001
#>                  Coverage of nominal 95% CI    0.9440 0.0073    0.9297    0.9583
#>   Bias corrected coverage of nominal 95% CI    0.9500 0.0069    0.9365    0.9635
#>                      Power of 5% level test    0.9740 0.0050    0.9641    0.9839
#> 
#>  Method = Exp, baseline = Weibull, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.3522     NA        NA        NA
#>                       Median point estimate   -0.3500     NA        NA        NA
#>                      Average standard error    0.0164     NA        NA        NA
#>                       Median standard error    0.0163     NA        NA        NA
#>                      Bias in point estimate    0.1478 0.0028    0.1423    0.1533
#>                    Empirical standard error    0.0886 0.0020    0.0847    0.0925
#>                          Mean squared error    0.0297 0.0009    0.0279    0.0314
#>  % gain in precision relative to method Cox    2.0701 0.0381    1.9954    2.1448
#>                  Model-based standard error    0.1280 0.0000    0.1279    0.1280
#>          Relative % error in standard error   44.4489 3.2309   38.1165   50.7813
#>                  Coverage of nominal 95% CI    0.8900 0.0099    0.8706    0.9094
#>   Bias corrected coverage of nominal 95% CI    0.9910 0.0030    0.9851    0.9969
#>                      Power of 5% level test    0.8870 0.0100    0.8674    0.9066
#> 
#>  Method = RP(2), baseline = Weibull, n = 250 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5072     NA        NA        NA
#>                       Median point estimate   -0.5055     NA        NA        NA
#>                      Average standard error    0.0171     NA        NA        NA
#>                       Median standard error    0.0171     NA        NA        NA
#>                      Bias in point estimate   -0.0072 0.0040   -0.0150    0.0007
#>                    Empirical standard error    0.1267 0.0028    0.1212    0.1323
#>                          Mean squared error    0.0161 0.0008    0.0146    0.0176
#>  % gain in precision relative to method Cox    1.0116 0.0067    0.9984    1.0248
#>                  Model-based standard error    0.1309 0.0001    0.1308    0.1310
#>          Relative % error in standard error    3.3058 2.3110   -1.2238    7.8353
#>                  Coverage of nominal 95% CI    0.9470 0.0071    0.9331    0.9609
#>   Bias corrected coverage of nominal 95% CI    0.9500 0.0069    0.9365    0.9635
#>                      Power of 5% level test    0.9740 0.0050    0.9641    0.9839
#> 
#>  Method = Cox, baseline = Exponential, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5255     NA        NA        NA
#>                       Median point estimate   -0.5226     NA        NA        NA
#>                      Average standard error    0.1016     NA        NA        NA
#>                       Median standard error    0.1003     NA        NA        NA
#>                      Bias in point estimate   -0.0255 0.0104   -0.0458   -0.0052
#>                    Empirical standard error    0.3277 0.0073    0.3133    0.3421
#>                          Mean squared error    0.1079 0.0047    0.0988    0.1170
#>  % gain in precision relative to method Cox    1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error    0.3188 0.0004    0.3180    0.3197
#>          Relative % error in standard error   -2.7072 2.1801   -6.9802    1.5658
#>                  Coverage of nominal 95% CI    0.9410 0.0075    0.9264    0.9556
#>   Bias corrected coverage of nominal 95% CI    0.9400 0.0075    0.9253    0.9547
#>                      Power of 5% level test    0.3740 0.0153    0.3440    0.4040
#> 
#>  Method = Exp, baseline = Exponential, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5223     NA        NA        NA
#>                       Median point estimate   -0.5181     NA        NA        NA
#>                      Average standard error    0.0977     NA        NA        NA
#>                       Median standard error    0.0970     NA        NA        NA
#>                      Bias in point estimate   -0.0223 0.0101   -0.0422   -0.0024
#>                    Empirical standard error    0.3208 0.0072    0.3068    0.3349
#>                          Mean squared error    0.1033 0.0045    0.0945    0.1122
#>  % gain in precision relative to method Cox    1.0431 0.0126    1.0183    1.0679
#>                  Model-based standard error    0.3126 0.0004    0.3118    0.3133
#>          Relative % error in standard error   -2.5836 2.1821   -6.8603    1.6932
#>                  Coverage of nominal 95% CI    0.9370 0.0077    0.9219    0.9521
#>   Bias corrected coverage of nominal 95% CI    0.9390 0.0076    0.9242    0.9538
#>                      Power of 5% level test    0.3850 0.0154    0.3548    0.4152
#> 
#>  Method = RP(2), baseline = Exponential, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5337     NA        NA        NA
#>                       Median point estimate   -0.5269     NA        NA        NA
#>                      Average standard error    0.1005     NA        NA        NA
#>                       Median standard error    0.0993     NA        NA        NA
#>                      Bias in point estimate   -0.0337 0.0105   -0.0543   -0.0131
#>                    Empirical standard error    0.3323 0.0074    0.3177    0.3468
#>                          Mean squared error    0.1114 0.0049    0.1018    0.1210
#>  % gain in precision relative to method Cox    0.9726 0.0066    0.9597    0.9855
#>                  Model-based standard error    0.3171 0.0004    0.3162    0.3179
#>          Relative % error in standard error   -4.5711 2.1383   -8.7620   -0.3802
#>                  Coverage of nominal 95% CI    0.9370 0.0077    0.9219    0.9521
#>   Bias corrected coverage of nominal 95% CI    0.9390 0.0076    0.9242    0.9538
#>                      Power of 5% level test    0.3920 0.0154    0.3617    0.4223
#> 
#>  Method = Cox, baseline = Weibull, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5123     NA        NA        NA
#>                       Median point estimate   -0.5133     NA        NA        NA
#>                      Average standard error    0.0926     NA        NA        NA
#>                       Median standard error    0.0900     NA        NA        NA
#>                      Bias in point estimate   -0.0123 0.0100   -0.0318    0.0073
#>                    Empirical standard error    0.3158 0.0071    0.3020    0.3297
#>                          Mean squared error    0.0998 0.0049    0.0903    0.1094
#>  % gain in precision relative to method Cox    1.0000 0.0000    1.0000    1.0000
#>                  Model-based standard error    0.3043 0.0005    0.3035    0.3052
#>          Relative % error in standard error   -3.6377 2.1601   -7.8714    0.5960
#>                  Coverage of nominal 95% CI    0.9450 0.0072    0.9309    0.9591
#>   Bias corrected coverage of nominal 95% CI    0.9440 0.0073    0.9297    0.9583
#>                      Power of 5% level test    0.3820 0.0154    0.3519    0.4121
#> 
#>  Method = Exp, baseline = Weibull, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.3492     NA        NA        NA
#>                       Median point estimate   -0.3506     NA        NA        NA
#>                      Average standard error    0.0834     NA        NA        NA
#>                       Median standard error    0.0825     NA        NA        NA
#>                      Bias in point estimate    0.1508 0.0066    0.1379    0.1636
#>                    Empirical standard error    0.2073 0.0046    0.1982    0.2163
#>                          Mean squared error    0.0656 0.0028    0.0602    0.0711
#>  % gain in precision relative to method Cox    2.3224 0.0503    2.2238    2.4209
#>                  Model-based standard error    0.2888 0.0002    0.2884    0.2891
#>          Relative % error in standard error   39.3334 3.1175   33.2231   45.4436
#>                  Coverage of nominal 95% CI    0.9740 0.0050    0.9641    0.9839
#>   Bias corrected coverage of nominal 95% CI    0.9950 0.0022    0.9906    0.9994
#>                      Power of 5% level test    0.1450 0.0111    0.1232    0.1668
#> 
#>  Method = RP(2), baseline = Weibull, n = 50 
#>                                              Estimate   MCSE Lower 95% Upper 95%
#>  Simulations with non-missing estimates/SEs 1000.0000     NA        NA        NA
#>                      Average point estimate   -0.5187     NA        NA        NA
#>                       Median point estimate   -0.5120     NA        NA        NA
#>                      Average standard error    0.0894     NA        NA        NA
#>                       Median standard error    0.0873     NA        NA        NA
#>                      Bias in point estimate   -0.0187 0.0100   -0.0382    0.0009
#>                    Empirical standard error    0.3155 0.0071    0.3017    0.3294
#>                          Mean squared error    0.0998 0.0047    0.0905    0.1091
#>  % gain in precision relative to method Cox    1.0020 0.0110    0.9804    1.0237
#>                  Model-based standard error    0.2989 0.0004    0.2982    0.2997
#>          Relative % error in standard error   -5.2531 2.1230   -9.4140   -1.0922
#>                  Coverage of nominal 95% CI    0.9380 0.0076    0.9231    0.9529
#>   Bias corrected coverage of nominal 95% CI    0.9430 0.0073    0.9286    0.9574
#>                      Power of 5% level test    0.4140 0.0156    0.3835    0.4445

Plotting summary statistics

Tables could get cumbersome when there are many different data-generating mechanisms. Plots are generally easier to interpret, and can be generated as easily as before.

Say we want to compare bias for each method by baseline hazard function and sample size using faceting:

ggplot(get_data(ss2, sstat = "bias"), aes(x = model, y = est, ymin = lower, ymax = upper)) +
  geom_hline(yintercept = 0, color = "red", lty = "dashed") +
  geom_point() +
  geom_errorbar(width = 1 / 3) +
  facet_grid(baseline ~ n) +
  theme_bw() +
  labs(x = "Method", y = "Bias")

References