Generating Tables of Descriptive Statistics

Motivation

R is well suited for statistical graphics, the application of advanced data analysis techniques, and Monte Carlo studies of estimators. However, it lacks support for the typical data management tasks as they arise in the social sciences as well as for the simple generation of desctiptive statistics. “memisc” facilitates not only typical data management tasks of survey researchers, but also the generation of descriptive statistics, as they are often a first step in serious social science data analysis. In particular it facilitates the creation of tables of percentages of other descriptive statistics broken down by subgroups in the data. This is mainly achieved by the function genTable, which is described in the following section. The section thereafter describes how tables thus created can be exported to LaTeX and HTML.

Note that these examples require data not included in the package (you need to register to GESIS to download the data). The vignette code cannot be run without this additional data.

Creating Tables of Descriptive Statistics

General table of descriptive statistics can be created using the function genTable(). The syntax of calls to this function is quite similar to that of the function xtabs(): The first argument (tagged formula) is a formula that determines the descriptive statistics used and by what groups they are computed. The left-hand side of the formula determines the statistics being computed. The right-hand side determines the grouping factor(s). The second argument is an optional data= argument that determines from which data frame or data set the descriptive statistics are to be computed. This is illustrated by the following example, which uses (like the page on item objects, see ?item) the GLES 2013 election study1. In this example we first create a table of some descriptives of the age distribution of the respondents per German federal state:

library(memisc)
ZA5702 <- spss.system.file("Data/ZA5702_v2-0-0.sav",
                           ignore.scale.info=TRUE) # Because the measurement info in the file is wrong.
gles2013work <- subset(ZA5702,
                       select=c(
                         wave                  = survey,
                         gender                = vn1,
                         byear                 = vn2c,
                         bmonth                = vn2b,
                         intent.turnout        = v10,
                         turnout               = n10,
                         voteint.candidate     = v11aa,
                         voteint.list          = v11ba,
                         postal.vote.candidate = v12aa,
                         postal.vote.list      = v12ba,
                         vote.candidate        = n11aa,
                         vote.list             = n11ba,                 
                         bula                  = bl
                       ))
gles2013work <- within(gles2013work,{
  measurement(byear) <- "interval"
  measurement(bmonth) <- "interval"
  age <- 2013 - byear
  age[bmonth > 9] <- age[bmonth > 9] - 1
  
})
options(digits=3)
age.tab <- genTable(c(Mean=mean(age),
           `Std.dev`=sd(age),
           Median=median(age))~bula,
         data=gles2013work)
age.tab
         bula
          Baden-Wuerttemberg Bayern Berlin Brandenburg Bremen Hamburg Hessen
  Mean                    55     54     53          60     60      51     57
  Std.dev                 19     19     20          19     12      19     19
  Median                  57     56     57          62     63      53     60
         bula
          Mecklenburg-Vorpommern Niedersachsen Nordrhein-Westfalen
  Mean                        57            55                  54
  Std.dev                     19            18                  19
  Median                      60            56                  55
         bula
          Rheinland-Pfalz Saarland Sachsen Sachsen-Anhalt Schleswig-Holstein
  Mean                 57       62      58             55                 60
  Std.dev              18       17      17             17                 20
  Median               60       65      60             56                 65
         bula
          Thueringen
  Mean            58
  Std.dev         17
  Median          60

This table does not look good, so we transprose it:

age.tab <- t(age.tab)
age.tab
bula                     Mean Std.dev Median
  Baden-Wuerttemberg     54.5    18.9   57.0
  Bayern                 54.4    18.9   56.0
  Berlin                 52.8    19.8   57.0
  Brandenburg            59.7    19.3   62.5
  Bremen                 60.4    11.5   63.0
  Hamburg                51.5    18.7   53.0
  Hessen                 56.9    18.5   60.0
  Mecklenburg-Vorpommern 57.0    19.2   60.5
  Niedersachsen          55.1    18.4   56.0
  Nordrhein-Westfalen    53.9    19.1   55.0
  Rheinland-Pfalz        57.2    18.2   60.5
  Saarland               61.9    17.3   65.0
  Sachsen                58.3    16.7   60.5
  Sachsen-Anhalt         54.7    17.1   56.0
  Schleswig-Holstein     60.0    19.9   65.0
  Thueringen             57.8    17.4   60.0

In the next example we create a table of percentages of the second votes per federal state. First we have to prepare the data, though:

gles2013work <- within(gles2013work,{

  candidate.vote <- cases(
              wave == 1 & intent.turnout == 6 -> postal.vote.candidate,
              wave == 1 & intent.turnout %in% 4:5 -> 900,
              wave == 1 & intent.turnout %in% 1:3 -> voteint.candidate,
              wave == 2 & turnout == 1 -> vote.candidate,
              wave == 2 & turnout == 2 -> 900
            )

  list.vote <- cases(
              wave == 1 & intent.turnout == 6 -> postal.vote.list,
              wave == 1 & intent.turnout %in% 4:5 -> 900,
              wave == 1 & intent.turnout %in% 1:3 -> voteint.list,
              wave == 2 & turnout ==1 -> vote.list,
              wave == 2 & turnout ==2 -> 900
            )

  candidate.vote <- recode(as.item(candidate.vote),
                      "CDU/CSU"   =  1 <- 1,
                      "SPD"       =  2 <- 4,
                      "FDP"       =  3 <- 5,
                      "Grüne"     =  4 <- 6,
                      "Linke"     =  5 <- 7,
                      "NPD"       =  6 <- 206,
                      "Piraten"   =  7 <- 215,
                      "AfD"       =  8 <- 322,
                      "Other"     = 10 <- 801,
                      "No Vote"   = 90 <- 900,
                      "WN"        = 98 <- -98,
                      "KA"        = 99 <- -99
                  )
  list.vote <- recode(as.item(list.vote),
                      "CDU/CSU"   =  1 <- 1,
                      "SPD"       =  2 <- 4,
                      "FDP"       =  3 <- 5,
                      "Grüne"     =  4 <- 6,
                      "Linke"     =  5 <- 7,
                      "NPD"       =  6 <- 206,
                      "Piraten"   =  7 <- 215,
                      "AfD"       =  8 <- 322,
                      "Other"     = 10 <- 801,
                      "No Vote"   = 90 <- 900,
                      "WN"        = 98 <- -98,
                      "KA"        = 99 <- -99
                  )
  
   missing.values(candidate.vote) <- 98:99
   missing.values(list.vote) <- 98:99
   measurement(candidate.vote) <- "nominal"
   measurement(list.vote) <- "nominal"
})
Warning messages:
1: In cases(postal.vote.candidate <- wave == 1 & intent.turnout ==  :
  78 NAs created
2: In cases(postal.vote.list <- wave == 1 & intent.turnout == 6, 900 <- wave ==  :
  78 NAs created
3: In recode(as.item(candidate.vote), `CDU/CSU` = 1 <- 1, SPD = 2 <- 4,  :
  recoding created 18 NAs
4: In recode(as.item(list.vote), `CDU/CSU` = 1 <- 1, SPD = 2 <- 4,  :
  recoding created 19 NAs

(When the code is run, some warnings are issued, that indicate that the conditions are not exhaustive, that is, there are some observations for which none of the conditions in the call cases() are met. The corresponding elements of resulting vector will contain NA for these observations. In the present case this occurs with observations that have missing values in both intent.turnout and turnout.)

After having set up the data, we get our table of percentages:

vote.tab <- genTable(percent(list.vote)~bula,
                     data=gles2013work)
options(digits=1)
t(vote.tab)
bula                     CDU/CSU SPD FDP Grüne Linke NPD Piraten AfD Other No Vote   N
  Baden-Wuerttemberg          28  22   7    17     6 0.4     2.1 4.6   1.1      12 285
  Bayern                      36  18   6    11     5 0.0     2.4 4.0   2.0      16 451
  Berlin                      27  22   8    10    14 1.8     1.8 6.6   0.6       8 166
  Brandenburg                 20  23   2     6    19 0.6     0.6 2.5   1.2      25 162
  Bremen                      22  26   0    17    13 0.0     0.0 4.3   0.0      17  23
  Hamburg                     22  36   2     4     7 2.2     0.0 4.4   2.2      20  45
  Hessen                      42  26   3     8     4 0.0     0.5 3.0   0.0      12 200
  Mecklenburg-Vorpommern      33  20   2     4    18 1.4     2.7 1.4   0.0      18 146
  Niedersachsen               33  32   3    10     3 0.0     0.7 0.7   0.4      17 284
  Nordrhein-Westfalen         33  31   3    11     4 0.4     2.3 1.8   0.7      13 563
  Rheinland-Pfalz             39  21   2     6     9 1.6     0.8 3.9   1.6      15 127
  Saarland                    40  40   0     0     0 0.0     0.0 0.0   0.0      20  30
  Sachsen                     49  17   1     3    14 0.3     1.2 0.9   0.3      13 332
  Sachsen-Anhalt              27  29   1     8    19 0.4     0.8 0.4   0.0      13 241
  Schleswig-Holstein          28  26   4     9     4 0.0     0.0 5.2   0.9      22 116
  Thueringen                  35  16   2     3    22 1.2     0.0 2.4   0.8      18 245

It is of course also possible to create multi-dimensional tables, i.e. tables created by grouping by more than one factor:

gles2013work <- within(gles2013work,{

  # We relabel the items, since they are originally in German
  labels(turnout) <- c("Yes, voted"=1, "No, did not vote"=2)   
  labels(gender) <- c("Male"=1,"Female"=2)
})
genTable(percent(turnout)~gender+bula,
         data=gles2013work)
, , bula = Baden-Wuerttemberg

                  gender
                   Male Female
  Yes, voted         88     85
  No, did not vote   12     15
  N                  90     61

, , bula = Bayern

                  gender
                   Male Female
  Yes, voted         85     80
  No, did not vote   15     20
  N                  89    129

, , bula = Berlin

                  gender
                   Male Female
  Yes, voted        100     85
  No, did not vote    0     15
  N                  38     52

, , bula = Brandenburg

                  gender
                   Male Female
  Yes, voted         83     77
  No, did not vote   17     23
  N                  36     62

, , bula = Bremen

                  gender
                   Male Female
  Yes, voted         91     80
  No, did not vote    9     20
  N                  11      5

, , bula = Hamburg

                  gender
                   Male Female
  Yes, voted         88     76
  No, did not vote   12     24
  N                  16     21

, , bula = Hessen

                  gender
                   Male Female
  Yes, voted         91     81
  No, did not vote    9     19
  N                  66     48

, , bula = Mecklenburg-Vorpommern

                  gender
                   Male Female
  Yes, voted         84     72
  No, did not vote   16     28
  N                  32     47

, , bula = Niedersachsen

                  gender
                   Male Female
  Yes, voted         88     83
  No, did not vote   12     17
  N                  75     70

, , bula = Nordrhein-Westfalen

                  gender
                   Male Female
  Yes, voted         90     82
  No, did not vote   10     18
  N                 148    158

, , bula = Rheinland-Pfalz

                  gender
                   Male Female
  Yes, voted         84     85
  No, did not vote   16     15
  N                  43     34

, , bula = Saarland

                  gender
                   Male Female
  Yes, voted         91     72
  No, did not vote    9     28
  N                  11     18

, , bula = Sachsen

                  gender
                   Male Female
  Yes, voted         88     88
  No, did not vote   12     12
  N                 103     73

, , bula = Sachsen-Anhalt

                  gender
                   Male Female
  Yes, voted         89     81
  No, did not vote   11     19
  N                  63     73

, , bula = Schleswig-Holstein

                  gender
                   Male Female
  Yes, voted         89     85
  No, did not vote   11     15
  N                  37     33

, , bula = Thueringen

                  gender
                   Male Female
  Yes, voted         91     71
  No, did not vote    9     29
  N                  70     73

Formatting Tables of Descriptive Statistics

The results of genTable() are objects of class "table" so that they can be re-arranged into a “flattened” table by the function ftable. To demonstrate this, we continue the previous example:

gt <- genTable(percent(turnout)~gender+bula,
         data=gles2013work)
# We beautify the table a bit ...
names(dimnames(gt)) <- c("Voted","Gender","State")
gt <- dimrename(gt,"Yes, voted"="Yes",
                "No, did not vote"="No")
ftable(gt,col.vars = c("Gender","Voted"))
                       Gender Male         Female        
                       Voted   Yes  No   N    Yes  No   N
State                                                    
Baden-Wuerttemberg              88  12  90     85  15  61
Bayern                          85  15  89     80  20 129
Berlin                         100   0  38     85  15  52
Brandenburg                     83  17  36     77  23  62
Bremen                          91   9  11     80  20   5
Hamburg                         88  12  16     76  24  21
Hessen                          91   9  66     81  19  48
Mecklenburg-Vorpommern          84  16  32     72  28  47
Niedersachsen                   88  12  75     83  17  70
Nordrhein-Westfalen             90  10 148     82  18 158
Rheinland-Pfalz                 84  16  43     85  15  34
Saarland                        91   9  11     72  28  18
Sachsen                         88  12 103     88  12  73
Sachsen-Anhalt                  89  11  63     81  19  73
Schleswig-Holstein              89  11  37     85  15  33
Thueringen                      91   9  70     71  29  73

Arranging the cells of a table using ftable() improves the appearance of the results of genTable() on screen, but to include the results into a word processor document or a LaTeX file, further facilities are needed and provided by “memisc”. To include the flattened table into a LaTeX document, one can convert and store it in the appropriate format using toLatex() and writeLines()

ft <- ftable(gt,col.vars = c("Gender","Voted"))
lt <- toLatex(ft,digits=c(1,1,0,1,1,0))
writeLines(lt,con="Voted2013-GenderState.tex")

For HTML output, one can use show_html() (e.g. for inclusion in “knitr” documents) and write_html(), both functions being based on format_html(). Here we continue the example to demonstate this:

show_html(ft,digits=c(1,1,0,1,1,0))
Gender:MaleFemale
StateVoted:YesNoNYesNoN
Baden-Wuerttemberg87.812.29085.214.861
Bayern85.414.68979.820.2129
Berlin100.00.03884.615.452
Brandenburg83.316.73677.422.662
Bremen90.99.11180.020.05
Hamburg87.512.51676.223.821
Hessen90.99.16681.218.848
Mecklenburg-Vorpommern84.415.63272.327.747
Niedersachsen88.012.07582.917.170
Nordrhein-Westfalen89.910.114882.317.7158
Rheinland-Pfalz83.716.34385.314.734
Saarland90.99.11172.227.818
Sachsen88.311.710387.712.373
Sachsen-Anhalt88.911.16380.819.273
Schleswig-Holstein89.210.83784.815.233
Thueringen91.48.67071.228.873
show_html(ft,digits=c(1,1,0,1,1,0),show.titles=FALSE)
MaleFemale
YesNoNYesNoN
Baden-Wuerttemberg87.812.29085.214.861
Bayern85.414.68979.820.2129
Berlin100.00.03884.615.452
Brandenburg83.316.73677.422.662
Bremen90.99.11180.020.05
Hamburg87.512.51676.223.821
Hessen90.99.16681.218.848
Mecklenburg-Vorpommern84.415.63272.327.747
Niedersachsen88.012.07582.917.170
Nordrhein-Westfalen89.910.114882.317.7158
Rheinland-Pfalz83.716.34385.314.734
Saarland90.99.11172.227.818
Sachsen88.311.710387.712.373
Sachsen-Anhalt88.911.16380.819.273
Schleswig-Holstein89.210.83784.815.233
Thueringen91.48.67071.228.873
# Writing into a HTML file ...
write_html(ft,digits=c(1,1,0,1,1,0),show.titles=FALSE,
           file="Voted2013-GenderState.html")

Continuing another example:

# age.tab was created earlier
age.ftab <- ftable(age.tab,row.vars=2)
show_html(age.ftab,digits=1,show.titles=FALSE)
MeanStd.devMedian
Baden-Wuerttemberg54.518.957.0
Bayern54.418.956.0
Berlin52.819.857.0
Brandenburg59.719.362.5
Bremen60.411.563.0
Hamburg51.518.753.0
Hessen56.918.560.0
Mecklenburg-Vorpommern57.019.260.5
Niedersachsen55.118.456.0
Nordrhein-Westfalen53.919.155.0
Rheinland-Pfalz57.218.260.5
Saarland61.917.365.0
Sachsen58.316.760.5
Sachsen-Anhalt54.717.156.0
Schleswig-Holstein60.019.965.0
Thueringen57.817.460.0

Of course we can also export to LaTeX:

toLatex(age.ftab,digits=1,show.titles=FALSE)
\begin{tabular}{llD{.}{.}{1}D{.}{.}{1}D{.}{.}{1}}
\toprule
 && \multicolumn{1}{c}{Mean}&\multicolumn{1}{c}{Std.dev}&\multicolumn{1}{c}{Median}\\
\midrule
Baden-Wuerttemberg     && 54.5 & 18.9 & 57.0\\
Bayern                 && 54.4 & 18.9 & 56.0\\
Berlin                 && 52.8 & 19.8 & 57.0\\
Brandenburg            && 59.7 & 19.3 & 62.5\\
Bremen                 && 60.4 & 11.5 & 63.0\\
Hamburg                && 51.5 & 18.7 & 53.0\\
Hessen                 && 56.9 & 18.5 & 60.0\\
Mecklenburg-Vorpommern && 57.0 & 19.2 & 60.5\\
Niedersachsen          && 55.1 & 18.4 & 56.0\\
Nordrhein-Westfalen    && 53.9 & 19.1 & 55.0\\
Rheinland-Pfalz        && 57.2 & 18.2 & 60.5\\
Saarland               && 61.9 & 17.3 & 65.0\\
Sachsen                && 58.3 & 16.7 & 60.5\\
Sachsen-Anhalt         && 54.7 & 17.1 & 56.0\\
Schleswig-Holstein     && 60.0 & 19.9 & 65.0\\
Thueringen             && 57.8 & 17.4 & 60.0\\
\bottomrule
\end{tabular}

  1. The German Longitudinal Election Study is funded by the German National Science Foundation (DFG) and carried out outin close cooperation with the DGfW, German Society for Electoral Studies. Principal investigators are Hans Rattinger (University of Mannheim, until 2014), Sigrid Roßteutscher (University of Frankfurt), Rüdiger Schmitt-Beck (University of Mannheim), Harald Schoen (Mannheim Centre for European Social Research, from 2015), Bernhard Weßels (Social Science Research Center Berlin), and Christof Wolf (GESIS – Leibniz Institute for the Social Sciences, since 2012). Neither the funding organisation nor the principal investigators bear any responsibility for the example code shown here.↩︎