search

TysonStanley / furniture

The furniture R package contains table1 for publication-ready simple and stratified descriptive statistics, tableC for publication-ready correlation matrixes, and other tables #rstats
49 stars 7 forks source link
cran descriptive-statistics exploratory-data-analysis health table-one table1 tables tidy tidyverse

readme

CRAN Status Badge Downloads R-CMD-check

furniture: v1.10.0

The furniture R package contains functions to help with data cleaning/tidying (e.g., washer(), rowmeans(), rowsums()), exploratory data analysis and reporting (e.g., table1(), tableC(), tableF()). It currently contains eight main functions:

  1. table1() : gives a well-formatted table for academic publication of descriptive statistics. Very useful for quick analyses as well. Notably, table1() now works with dplyr::group_by().
  2. tableC() : gives a well-formatted table of correlations.
  3. tableF() : provides a thorough frequency table for quick checks of the levels of a variable.
  4. washer() : changes several values in a variable (very useful for changing place holder values to missing).
  5. long() : is a wrapper of stats::reshape(), takes the data from wide to long format (long is often the tidy version of the data), works well with the tidyverse, and can handle unbalanced multilevel data.
  6. wide() : also a wrapper of stats::reshape(), takes the data from long to wide, and like long(), works well with the tidyverse and can handle unbalanced multilevel data.
  7. rowmeans() and rowmeans.n() : tidyverse friendly versions of rowMeans(), where the rowmeans.n() function allows n number of missing
  8. rowsums() and rowsums.n() : tidyverse friendly versions of rowSums(), where the rowsums.n() function allows n number of missing

In conjunction with many other tidy tools, the package should be useful for health, behavioral, and social scientists working on quantitative research.

Installation

The latest stable build of the package can be downloaded from CRAN via:

install.packages("furniture")

You can download the developmental version via:

remotes::install_github("tysonstanley/furniture")

Using furniture

The main functions are the table*() functions (e.g., table1(), tableC(), tableF()).

library(furniture)
data("nhanes_2010")

table1(nhanes_2010,
       age, marijuana, illicit, rehab,
       splitby=~asthma,
       na.rm = FALSE)
#> 
#> ───────────────────────────────────
#>                   asthma 
#>            Yes         No         
#>            n = 251     n = 1164   
#>  age                              
#>            23.0 (3.9)  23.4 (4.0) 
#>  marijuana                        
#>     Yes    131 (52.2%) 584 (50.2%)
#>     No     97 (38.6%)  434 (37.3%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  illicit                          
#>     Yes    23 (9.2%)   117 (10.1%)
#>     No     205 (81.7%) 901 (77.4%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  rehab                            
#>     Yes    10 (4%)     37 (3.2%)  
#>     No     121 (48.2%) 547 (47%)  
#>     NA     120 (47.8%) 580 (49.8%)
#> ───────────────────────────────────
table1(nhanes_2010,
       age, marijuana, illicit, rehab,
       splitby=~asthma, 
       output = "text2",
       na.rm = FALSE)
#> 
#> ───────────────────────────────────
#>                   asthma 
#>            Yes         No         
#>            n = 251     n = 1164   
#>  --------- ----------- -----------
#>  age                              
#>            23.0 (3.9)  23.4 (4.0) 
#>  marijuana                        
#>     Yes    131 (52.2%) 584 (50.2%)
#>     No     97 (38.6%)  434 (37.3%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  illicit                          
#>     Yes    23 (9.2%)   117 (10.1%)
#>     No     205 (81.7%) 901 (77.4%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  rehab                            
#>     Yes    10 (4%)     37 (3.2%)  
#>     No     121 (48.2%) 547 (47%)  
#>     NA     120 (47.8%) 580 (49.8%)
#> ───────────────────────────────────
library(tidyverse)
nhanes_2010 %>%
  group_by(asthma) %>%
  table1(age, marijuana, illicit, rehab,
         output = "text2",
         na.rm = FALSE)
#> 
#> ───────────────────────────────────
#>                   asthma 
#>            Yes         No         
#>            n = 251     n = 1164   
#>  --------- ----------- -----------
#>  age                              
#>            23.0 (3.9)  23.4 (4.0) 
#>  marijuana                        
#>     Yes    131 (52.2%) 584 (50.2%)
#>     No     97 (38.6%)  434 (37.3%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  illicit                          
#>     Yes    23 (9.2%)   117 (10.1%)
#>     No     205 (81.7%) 901 (77.4%)
#>     NA     23 (9.2%)   146 (12.5%)
#>  rehab                            
#>     Yes    10 (4%)     37 (3.2%)  
#>     No     121 (48.2%) 547 (47%)  
#>     NA     120 (47.8%) 580 (49.8%)
#> ───────────────────────────────────

table1() can do bivariate significance tests as well.

library(tidyverse)
nhanes_2010 %>%
  group_by(asthma) %>%
  table1(age, marijuana, illicit, rehab,
         output = "text2",
         na.rm = FALSE,
         test = TRUE)
#> 
#> ───────────────────────────────────────────
#>                   asthma 
#>            Yes         No          P-Value
#>            n = 251     n = 1164           
#>  --------- ----------- ----------- -------
#>  age                               0.201  
#>            23.0 (3.9)  23.4 (4.0)         
#>  marijuana                         1      
#>     Yes    131 (52.2%) 584 (50.2%)        
#>     No     97 (38.6%)  434 (37.3%)        
#>     NA     23 (9.2%)   146 (12.5%)        
#>  illicit                           0.623  
#>     Yes    23 (9.2%)   117 (10.1%)        
#>     No     205 (81.7%) 901 (77.4%)        
#>     NA     23 (9.2%)   146 (12.5%)        
#>  rehab                             0.729  
#>     Yes    10 (4%)     37 (3.2%)          
#>     No     121 (48.2%) 547 (47%)          
#>     NA     120 (47.8%) 580 (49.8%)        
#> ───────────────────────────────────────────

By default it does the appropriate parametric tests. However, you can change that by setting param = FALSE (new with v 1.9.1).

library(tidyverse)
nhanes_2010 %>%
  group_by(asthma) %>%
  table1(age, marijuana, illicit, rehab,
         output = "text2",
         na.rm = FALSE,
         test = TRUE,
         param = FALSE,
         type = "condense")
#> 
#> ───────────────────────────────────────────────
#>                       asthma 
#>                Yes         No          P-Value
#>                n = 251     n = 1164           
#>  ------------- ----------- ----------- -------
#>  age           23.0 (3.9)  23.4 (4.0)  0.235  
#>  marijuana: No 97 (38.6%)  434 (37.3%) 1      
#>  illicit: No   205 (81.7%) 901 (77.4%) 0.623  
#>  rehab: No     121 (48.2%) 547 (47%)   0.729  
#> ───────────────────────────────────────────────

It can also do a total column with the stratified columns (new with v 1.9.0) with the total = TRUE argument.

nhanes_2010 %>%
  group_by(asthma) %>%
  table1(age, marijuana, illicit, rehab,
         output = "text2",
         na.rm = FALSE,
         test = TRUE,
         type = "condense",
         total = TRUE)
#> 
#> ────────────────────────────────────────────────────────────
#>                                   asthma 
#>                Total        Yes         No          P-Value
#>                n = 1417     n = 251     n = 1164           
#>  ------------- ------------ ----------- ----------- -------
#>  age           23.3 (4.0)   23.0 (3.9)  23.4 (4.0)  0.201  
#>  marijuana: No 532 (37.5%)  97 (38.6%)  434 (37.3%) 1      
#>  illicit: No   1107 (78.1%) 205 (81.7%) 901 (77.4%) 0.623  
#>  rehab: No     668 (47.1%)  121 (48.2%) 547 (47%)   0.729  
#> ────────────────────────────────────────────────────────────

It can also report the statistics in addition to the p-values.

nhanes_2010 %>%
  group_by(asthma) %>%
  table1(age, marijuana, illicit, rehab,
         output = "text2",
         na.rm = FALSE,
         test = TRUE,
         total = TRUE,
         type = "full")
#> 
#> ─────────────────────────────────────────────────────────────────────────
#>                                       asthma 
#>            Total        Yes         No          Test             P-Value
#>            n = 1417     n = 251     n = 1164                            
#>  --------- ------------ ----------- ----------- ---------------- -------
#>  age                                            T-Test: -1.28    0.201  
#>            23.3 (4.0)   23.0 (3.9)  23.4 (4.0)                          
#>  marijuana                                      Chi Square: 0    1      
#>     Yes    716 (50.5%)  131 (52.2%) 584 (50.2%)                         
#>     No     532 (37.5%)  97 (38.6%)  434 (37.3%)                         
#>     NA     169 (11.9%)  23 (9.2%)   146 (12.5%)                         
#>  illicit                                        Chi Square: 0.24 0.623  
#>     Yes    141 (10%)    23 (9.2%)   117 (10.1%)                         
#>     No     1107 (78.1%) 205 (81.7%) 901 (77.4%)                         
#>     NA     169 (11.9%)  23 (9.2%)   146 (12.5%)                         
#>  rehab                                          Chi Square: 0.12 0.729  
#>     Yes    48 (3.4%)    10 (4%)     37 (3.2%)                           
#>     No     668 (47.1%)  121 (48.2%) 547 (47%)                           
#>     NA     701 (49.5%)  120 (47.8%) 580 (49.8%)                         
#> ─────────────────────────────────────────────────────────────────────────

table1() can be outputted directly to other formats. All knitr::kable() options are available for this and there is an extra option "latex2" which provides a publication ready table in Latex documents.

A new function table1_gt() integrates the fantastic gt package to make beautiful HTML tables for table1.

nhanes_2010 %>% 
  group_by(asthma) %>% 
  table1(age, marijuana, illicit, rehab, na.rm = FALSE) %>% 
  table1_gt() %>% 
  print()
#> Using dplyr::group_by() groups: asthma

<div id="uboagfutpu" style="padding-left:0px;padding-right:0px;padding-top:10px;padding-bottom:10px;overflow-x:auto;overflow-y:auto;width:auto;height:auto;">

Characteristic Yes, n = 251 No, n = 1164

age

23.0 (3.9)

23.4 (4.0)

marijuana

     Yes

131 (52.2%)

584 (50.2%)

     No

97 (38.6%)

434 (37.3%)

     NA

23 (9.2%)

146 (12.5%)

illicit

     Yes

23 (9.2%)

117 (10.1%)

     No

205 (81.7%)

901 (77.4%)

     NA

23 (9.2%)

146 (12.5%)

rehab

     Yes

10 (4%)

37 (3.2%)

     No

121 (48.2%)

547 (47%)

     NA

120 (47.8%)

580 (49.8%)

The tableC() function gives a well-formatted correlation table.

tableC(nhanes_2010, 
       age, active, vig_active, 
       na.rm=TRUE)
#> N = 317
#> Note: pearson correlation (p-value).
#> 
#> ──────────────────────────────────────────────────
#>                [1]            [2]           [3]  
#>  [1]age        1.00                              
#>  [2]active     -0.148 (0.008) 1.00               
#>  [3]vig_active -0.083 (0.141) 0.828 (<.001) 1.00 
#> ──────────────────────────────────────────────────

The tableF() function gives a table of frequencies.

tableF(nhanes_2010, age)
#> 
#> ──────────────────────────────────
#>  age Freq CumFreq Percent CumPerc
#>  18  191  191     13.48%  13.48% 
#>  19  153  344     10.80%  24.28% 
#>  20  111  455     7.83%   32.11% 
#>  21  95   550     6.70%   38.81% 
#>  22  100  650     7.06%   45.87% 
#>  23  112  762     7.90%   53.78% 
#>  24  93   855     6.56%   60.34% 
#>  25  100  955     7.06%   67.40% 
#>  26  91   1046    6.42%   73.82% 
#>  27  77   1123    5.43%   79.25% 
#>  28  91   1214    6.42%   85.67% 
#>  29  86   1300    6.07%   91.74% 
#>  30  117  1417    8.26%   100.00%
#> ──────────────────────────────────

In addition, the rowmeans() and rowsums() functions offer a simplified use of rowMeans() and rowSums(), particularly when using the tidyverse’s mutate().

nhanes_2010 %>%
  select(vig_active, mod_active) %>%
  mutate(avg_active = rowmeans(vig_active, mod_active, na.rm=TRUE)) %>%
  mutate(sum_active = rowsums(vig_active, mod_active, na.rm=TRUE)) %>% 
  head()
#>   vig_active mod_active avg_active sum_active
#> 1         30         NA         30         30
#> 2        180        180        180        360
#> 3         NA         NA        NaN          0
#> 4         20         70         45         90
#> 5        120         NA        120        120
#> 6         NA         NA        NaN          0

The rowmeans.n() and rowsums.n() allow n missing values while still calculating the mean or sum.

df <- data.frame(
  x = c(NA, 1:5),
  y = c(1:5, NA)
)

df %>%
  mutate(avg = rowmeans.n(x, y, n = 1))
#>    x  y avg
#> 1 NA  1 1.0
#> 2  1  2 1.5
#> 3  2  3 2.5
#> 4  3  4 3.5
#> 5  4  5 4.5
#> 6  5 NA 5.0

Notes

The package is most useful in conjunction with other tidy tools to get data cleaned/tidied and start exploratory data analysis. I recommend using packages such as library(dplyr), library(tidyr), and library(ggplot2) with library(furniture) to accomplish this.

The original function–table1()–is simply built for both exploratory descriptive analysis and communication of findings. See vignettes or tysonbarrett.com for several examples of its use. Also see our paper in the R Journal.