This report was produced using the distill R package, which is intended for producing academic articles in html format. You can see that distill does not always interact well with user-produced html code.

Introduction

This is a set of reproducible examples for the R (R Development Team 2020) Hmisc package (Harrell 2020), put together in an rmarkdown html document using RStudio and knitr. When viewing the resulting html file you can see all the code, and there exist options to download the entire rmarkdown script, which is especially helpful for seeing how knitr chunks are specified. Graphics that have a plotly method for them are rendered using plotly instead of using defaults such as base graphics, lattice, or ggplot2. That way the plots are somewhat interactive, e.g., allow for drill-down for more information without the viewer needing to install R.

Much of the tabular output produced here was created using html methods, which are especially easy to implement with rmarkdown and make for output that can be directly opened using word processors. Jump to Computing Environment for a list of packages used here, and their version numbers.

Rmarkdown themes such as bookdown and rmdformats::readthedown (the latter is used here) allow one to number figures and the symbolically reference them. The knitrSet capfile='captions.md' argument below capitalizes on this to make it easy for the user to insert a table of figures anywhere in the report. Here we place it at the end. A caption listed in the table of figures is the short caption (scap= or fig.scap in the chunk header) if there is one, otherwise the long caption is used. If neither caption is used, that figure will not appear in the table of figures.

The full script for this report may be found here.

Setup

Fetching Data, Modifying Variables, and Printing Data Dictionary

The getHdata function is used to fetch a dataset from the Vanderbilt DataSets web site hbiostat.org/data. The upData function is used to

create a new variable from an old one
add labels to 2 variables
add units to the new variable
remove the old variable
automatically move units of measurements from parenthetical expressions in labels to separate units attributed used by Hmisc and rms functions for table making and graphics

contents is used to print a data dictionary, run through an html method for nicer output. Information about the data source may be found here. Click on the number of levels in the contents table to jump to the value labels for the variable.

getHdata(pbc)

# Have upData move units from labels to separate attribute
pbc <- upData(pbc,
              fu.yrs = fu.days / 365.25,
              labels = c(fu.yrs = 'Follow-up Time',
                         status = 'Death or Liver Transplantation'),
              units = c(fu.yrs = 'year'),
              drop  = 'fu.days',
              moveUnits=TRUE, html=TRUE)

Input object size:   80952 bytes;    19 variables    418 observations
 Label for bili changed from Serum Bilirubin (mg/dl) to Serum Bilirubin 
    units set to mg/dl 
 Label for albumin changed from Albumin (gm/dl) to Albumin 
    units set to gm/dl 
 Label for protime changed from Prothrombin Time (sec.) to Prothrombin Time 
    units set to sec. 
 Label for alk.phos changed from Alkaline Phosphatase (U/liter) to Alkaline Phosphatase 
    units set to U/liter 
 Label for sgot changed from SGOT (U/ml) to SGOT 
    units set to U/ml 
 Label for chol changed from Cholesterol (mg/dl) to Cholesterol 
    units set to mg/dl 
 Label for trig changed from Triglycerides (mg/dl) to Triglycerides 
    units set to mg/dl 
 Label for platelet changed from Platelets (per cm^3/1000) to Platelets 
    units set to per cm^3/1000 
 Label for copper changed from Urine Copper (ug/day) to Urine Copper 
    units set to ug/day 
 Added variable     fu.yrs
 Dropped variable   fu.days
 New object size:   84744 bytes;    19 variables    418 observations

# The following can also be done by running this command
# to put the results in a new browser tab:
# getHdata(pbc, 'contents')
html(contents(pbc), maxlevels=10, levelType='table')

Data frame:pbc

418 observations and 19 variables, maximum # NAs:136

Name	Labels	Units	Levels	Storage	NAs
bili	Serum Bilirubin	mg/dl		double	0
albumin	Albumin	gm/dl		double	0
stage	Histologic Stage, Ludwig Criteria			integer	6
protime	Prothrombin Time	sec.		double	2
sex			2	integer	0
age	Age			double	0
spiders			2	integer	106
hepatom			2	integer	106
ascites			2	integer	106
alk.phos	Alkaline Phosphatase	U/liter		double	106
sgot	SGOT	U/ml		double	106
chol	Cholesterol	mg/dl		integer	134
trig	Triglycerides	mg/dl		integer	136
platelet	Platelets	per cm^3/1000		integer	110
drug			3	integer	0
status	Death or Liver Transplantation			integer	0
edema			3	integer	0
copper	Urine Copper	ug/day		integer	108
fu.yrs	Follow-up Time	year		double	0

Variable	Levels
sex	male
	female
spiders, hepatom	absent
ascites	present
drug	D-penicillamine
	placebo
	not randomized
edema	no edema
	edema, no diuretic therapy
	edema despite diuretic therapy

Descriptive Statistics Without Stratification

The html method is used for the describe function, and the output is put in a scrollable box. Other than for the overall title and variable names and labels, the output size used here is 80 (0.8 × the usual font size¹). But the graphical display of the descriptives statistics that follows this is probably better.

# did have results='asis' above
d <- describe(pbc)
html(d, size=80, scroll=TRUE)

pbc

19 Variables 418 Observations

bili: Serum Bilirubin mg/dl

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
418	0	98	0.998	3.221	3.742	0.50	0.60	0.80	1.40	3.40	8.03	14.00

lowest : 0.3 0.4 0.5 0.6 0.7 , highest: 21.6 22.5 24.5 25.5 28.0

albumin: Albumin gm/dl

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
418	0	154	1	3.497	0.473	2.750	2.967	3.243	3.530	3.770	4.010	4.141

lowest : 1.96 2.10 2.23 2.27 2.31 , highest: 4.30 4.38 4.40 4.52 4.64

stage: Histologic Stage, Ludwig Criteria

n	missing	distinct	Info	Mean	Gmd
412	6	4	0.893	3.024	0.9519

 Value          1     2     3     4
 Frequency     21    92   155   144
 Proportion 0.051 0.223 0.376 0.350

protime: Prothrombin Time sec.

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
416	2	48	0.998	10.73	1.029	9.60	9.80	10.00	10.60	11.10	12.00	12.45

lowest : 9.0 9.1 9.2 9.3 9.4 , highest: 13.8 14.1 15.2 17.1 18.0

sex

n	missing	distinct
418	0	2

 Value        male female
 Frequency      44    374
 Proportion  0.105  0.895

age: Age

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
418	0	345	1	50.74	11.96	33.84	36.37	42.83	51.00	58.24	64.30	67.92

lowest : 26.27789 28.88433 29.55510 30.27515 30.57358 , highest: 74.52430 75.00000 75.01164 76.70910 78.43943

spiders

n	missing	distinct
312	106	2

 Value       absent present
 Frequency      222      90
 Proportion   0.712   0.288

hepatom

n	missing	distinct
312	106	2

 Value       absent present
 Frequency      152     160
 Proportion   0.487   0.513

ascites

n	missing	distinct
312	106	2

 Value       absent present
 Frequency      288      24
 Proportion   0.923   0.077

alk.phos: Alkaline Phosphatase U/liter

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
312	106	295	1	1983	1760	599.6	663.0	871.5	1259.0	1980.0	3826.4	6669.9

lowest : 289.0 310.0 369.0 377.0 414.0 , highest: 11046.6 11320.2 11552.0 12258.8 13862.4

sgot: SGOT U/ml

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
312	106	179	1	122.6	60.45	54.25	60.45	80.60	114.70	151.90	196.47	219.25

lowest : 26.35 28.38 41.85 43.40 45.00 , highest: 288.00 299.15 328.60 338.00 457.25

chol: Cholesterol mg/dl

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
284	134	201	1	369.5	194.5	188.4	213.6	249.5	309.5	400.0	560.8	674.0

lowest : 120 127 132 149 151 , highest: 1336 1480 1600 1712 1775

trig: Triglycerides mg/dl

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
282	136	146	1	124.7	64.07	56.00	63.10	84.25	108.00	151.00	195.00	230.95

lowest : 33 44 46 49 50 , highest: 319 322 382 432 598

platelet: Platelets per cm³/1000

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
308	110	210	1	261.9	107.8	117.7	139.7	199.8	257.0	322.5	386.5	430.6

lowest : 62 70 71 79 80 , highest: 493 514 518 539 563

drug

n	missing	distinct
418	0	3

 Value      D-penicillamine         placebo  not randomized
 Frequency              154             158             106
 Proportion           0.368           0.378           0.254

status: Death or Liver Transplantation

n	missing	distinct	Info	Sum	Mean	Gmd
418	0	2	0.71	161	0.3852	0.4748

edema

n	missing	distinct
418	0	3

 Value                            no edema     edema, no diuretic therapy
 Frequency                             354                             44
 Proportion                          0.847                          0.105
                                          
 Value      edema despite diuretic therapy
 Frequency                              20
 Proportion                          0.048

copper: Urine Copper ug/day

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
310	108	158	1	97.65	83.16	17.45	24.00	41.25	73.00	123.00	208.10	249.20

lowest : 4 9 10 11 12 , highest: 412 444 464 558 588

fu.yrs: Follow-up Time year

n	missing	distinct	Info	Mean	Gmd	.05	.10	.25	.50	.75	.90	.95
418	0	399	1	5.251	3.429	0.671	1.661	2.992	4.736	7.155	9.649	11.063

lowest :	0.1122519	0.1177276	0.1396304	0.1943874	0.2108145
highest:	12.3203285	12.3449692	12.3832991	12.4736482	13.1279945

# prList is in Hmisc; useful for plotting or printing a list of objects
# Can just use plot(d) if don't care about the mess
# If using html output these 2 images would not be rendered no matter what
p <- plot(d)
# The option htmlfig=2 causes markupSpecs$html$cap() to be used to
# HTML-typeset as a figure caption and to put the sub-sub section
# marker ### in front of the caption.  htmlfig is the only reason
# results='asis' was needed in the chunk header
# We define a long caption for one of the plots, which does not appear
# in the table of contents
# prList works for html notebooks but not html documents
# prList(p, lcap=c('', 'These are spike histograms'), htmlfig=2)

You can also re-form multiple plotly graphs into a single HTML object. If you want to have total control over long and short figure captions, use the Hmisc putHfig function to render the result, with a caption and a short caption for the table of contents. That would have fixed a problem with the chunk below: when plotly graphics are not rendered in the usual way, the figure is not numbered and no caption appears.

htmltools::tagList(p)    # lapply(p, plotly::as.widget)

You can also create figure captions outside of R code by using the smg, fcap HTML tags defined in markupSpecs. The long caption not appearing in the table of contents will be in a separate line without ###.

Stratified Descriptive Statistics

Produce stratified quantiles, means/SD, and proportions by treatment group. Plot the results before rendering as an advanced html table:

categorical variables: a single dot chart
continuous variables: a series of extended box plots

s <- summaryM(bili + albumin + stage + protime + sex + age + spiders +
              alk.phos + sgot + chol ~ drug, data=pbc,
              overall=FALSE, test=TRUE)
plot(s, which='categorical')

Figure 1: Proportions and χ² tests for categorical variables

To construct the caption outside of the code chunk use e.g. ### r cap(‘Proportions and,’ mu$chisq(), ‘tests for categorical variables’) where a backtick is placed before r and after the last ).

plot(s, which='continuous', vars=1 : 4)

Figure 2: Extended box plots for the first 4 continuous variables

plot(s, which='continuous', vars=5 : 7)

Figure 3: Extended box plots for the remaining continuous variables

html(s, caption='Baseline characteristics by randomized treatment',
     exclude1=TRUE, npct='both', digits=3, middle.bold=TRUE,
     prmsd=TRUE, brmsd=TRUE, msdsize=mu$smaller2)

	N	D-penicillamine N=154	placebo N=158	not randomized N=106	Test Statistic
Baseline characteristics by randomized treatment.
Serum Bilirubin mg/dl	418	0.725 1.300 3.600 3.649 ± 5.282	0.800 1.400 3.200 2.873 ± 3.629	0.725 1.400 3.075 3.117 ± 4.043	F_2 415=0.03, P=0.972¹
Albumin gm/dl	418	3.342 3.545 3.777 3.524 ± 0.396	3.212 3.565 3.830 3.516 ± 0.443	3.125 3.470 3.720 3.431 ± 0.435	F_2 415=2.13, P=0.12¹
Histologic Stage, Ludwig Criteria : 1	412	0.03 ⁴⁄₁₅₄	0.08 ¹²⁄₁₅₈	0.05 ⁵⁄₁₀₀	χ²₆=5.33, P=0.502²
2		0.21 ³²⁄₁₅₄	0.22 ³⁵⁄₁₅₈	0.25 ²⁵⁄₁₀₀
3		0.42 ⁶⁴⁄₁₅₄	0.35 ⁵⁶⁄₁₅₈	0.35 ³⁵⁄₁₀₀
4		0.35 ⁵⁴⁄₁₅₄	0.35 ⁵⁵⁄₁₅₈	0.35 ³⁵⁄₁₀₀
Prothrombin Time sec.	416	10.000 10.600 11.400 10.800 ± 1.138	10.025 10.600 11.000 10.653 ± 0.851	10.100 10.600 11.000 10.750 ± 1.078	F_2 413=0.23, P=0.795¹
sex : female	418	0.90 ¹³⁹⁄₁₅₄	0.87 ¹³⁷⁄₁₅₈	0.92 ⁹⁸⁄₁₀₆	χ²₂=2.38, P=0.304²
Age	418	41.43 48.11 55.80 48.58 ± 9.96	42.98 51.93 58.90 51.42 ± 11.01	46.00 53.00 61.00 52.87 ± 9.78	F_2 415=6.1, P=0.002¹
spiders	312	0.29 ⁴⁵⁄₁₅₄	0.28 ⁴⁵⁄₁₅₈		χ²₁=0.02, P=0.885²
Alkaline Phosphatase U/liter	312	922 1283 1950 1943 ± 2102	841 1214 2028 2021 ± 2183		F_1 310=0.06, P=0.812³
SGOT U/ml	312	83.8 117.4 151.9 125.0 ± 58.9	76.7 111.6 151.5 120.2 ± 54.5		F_1 310=0.55, P=0.46³
Cholesterol mg/dl	284	254 304 377 374 ± 252	248 316 417 365 ± 210		F_1 282=0.37, P=0.545³
a b c represent the lower quartile a, the median b, and the upper quartile c for continuous variables. x ± s represents X ± 1 SD. N is the number of non-missing values. Tests used: ¹Kruskal-Wallis test; ²Pearson test; ³Wilcoxon test .

Now show almost the full raw data for one continuous variable stratified by treatment. This display spike histograms using at most 100 bins, and also shows the mean and quantiles similar to what is in an extended box plot: 0.05, 0.25, 0.5, 0.75, 0.95. Measures of spread are also shown if the user clicks on their legend entries: Gini’s mean difference (mean absolute difference between any two values) and the SD. These can be seen as horizontal lines up against the minimum x-value.

with(pbc, histboxp(x=sgot, group=drug, sd=TRUE))

Figure 4: Spike histograms, means, quantiles, Gini’s mean difference, and SD stratified by treatment

The following is a better way to display proportions, for categorical variables. If computing marginal statistics by running the dataset through the Hmisc addMarginal function, the plot method with options(grType='plotly') is especially useful.

pbcm <- addMarginal(pbc, drug)
s <- summaryP(stage + sex + spiders ~ drug, data=pbc)
# putHcap('Proportions stratified by treatment')
plot(s, groups='drug')

Figure 5: Proportions (large symbols) and proportions stratified by treatment (small symbols)

s <- summaryP(stage + sex + spiders ~ drug, data=pbcm)
plot(s, marginVal='All', marginLabel='All Treatment Groups')

Figure 5: Proportions (large symbols) and proportions stratified by treatment (small symbols)

Better Demonstration of Boxplot Replacement

getHdata(support2)
with(support2, histboxp(x=meanbp, group=dzgroup, sd=TRUE, bins=200))

Figure 6: Spike histograms, means, quantiles, Gini’s mean difference, and SD for MAP stratified by diagnosis

Changing Size of Figure Captions

As explained here, one can place captions under figures using ordinary knitr capabilities, and one can change the size of captions. The following example defines a CSS style to make captions small (here 0.6em), and produces a plot with a caption. Unlike using putHfig captions given in knitr chunks do not also appear in the table of contents.

# Note: in the chunk header cap is an alias for fig.cap defined by knitrSet
plot(runif(10))

Figure 7: This is a simple figure caption

Computing Environment²

 R version 4.0.3 (2020-10-10)
 Platform: x86_64-pc-linux-gnu (64-bit)
 Running under: Pop!_OS 20.10
 
 Matrix products: default
 BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.9.0
 LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.9.0
 
 attached base packages:
 [1] stats     graphics  grDevices utils     datasets  methods   base     
 
 other attached packages:
 [1] Hmisc_4.4-2     ggplot2_3.3.2   Formula_1.2-4   survival_3.2-7 
 [5] lattice_0.20-41

To cite R in publication use:

R Core Team (2020). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.

Bibliographic File Managament

Find and Install `.csl` Reference Style Files

# Note: mu was defined in an earlier code chunk
# Only need to install .csl file once.
mu$installcsl(rec=TRUE)   # get list of recommended styles
mu$installcsl()     # web search of styles meeting your criteria
# Install a .csl file to your project directory:
mu$installcsl('american-medical-association')

List of Figures

Figure	Description
??	Two `plotly` graphics combined into one
1	Proportions and χ² tests for categorical variables
2	Extended box plots for the first 4 continuous variables
3	Extended box plots for the remaining continuous variables
4	Stratified spike histograms and quantiles
5	Proportions with and without stratification by treatment
6	Stratified spike histograms and quantiles for MAP
7	This is a simple figure caption

Note: the hidden R command that rendered the table of figures (including short captions) was markupSpecs$markdown$tof().

Harrell, Frank E. 2020. “Hmisc: A Package of Miscellaneous R Functions.” https://hbiostat.org/R/Hmisc.

R Development Team. 2020. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; www.r-project.org. http://www.R-project.org.

The default is 75% size↩︎
mu is a copy of the part of the Hmisc package object markupSpecs that is for html. It includes a function session that renders the session environment (including package versions) in html.↩︎

Hmisc

Introduction

Setup

Fetching Data, Modifying Variables, and Printing Data Dictionary

Data frame:pbc

Descriptive Statistics Without Stratification

Stratified Descriptive Statistics

Better Demonstration of Boxplot Replacement

Changing Size of Figure Captions

Computing Environment2

Bibliographic File Managament

Find and Install .csl Reference Style Files

List of Figures

References

Computing Environment²

Find and Install `.csl` Reference Style Files