Does anyone know of an R equivalent to SAS PROC FREQ?
I am trying to generate summary descriptive statistics for multiple variables at once.
$\begingroup$
$\endgroup$
-
2$\begingroup$ Why was this question closed? It relates to data visualization and generated several worthwhile responses. $\endgroup$ – z0lo Feb 28 '13 at 14:26
$\begingroup$
$\endgroup$
I use table and prop.table, but CrossTable in the gmodels package might give you results even closer to SAS. See this link.
Also, to generate "descriptive statistics for multiple variables at once," you would use the summary function; e.g., summary(mydata).
-
$\begingroup$ In addition, I warmly recommend the vcd package, but see the accompanying vignette: Working with categorical data with R and the vcd and vcdExtra packages. $\endgroup$ – chl Aug 4 '11 at 19:19
$\begingroup$
$\endgroup$
Summarising data in base R is just a headache. This is one of the areas where SAS works quite well. For R, I recommend the plyr package.
In SAS:
/* tabulate by a and b, with summary stats for x and y in each cell */
proc summary data=dat nway;
class a b;
var x y;
output out=smry mean(x)=xmean mean(y)=ymean var(y)=yvar;
run;
with plyr:
smry <- ddply(dat, .(a, b), summarise, xmean=mean(x), ymean=mean(y), yvar=var(y))
$\begingroup$
$\endgroup$
I don't use SAS; so I can't comment on whether the following replicate SAS PROC FREQ, but these are two quick strategies for describing variables in a data.frame that I often use:
describeinHmiscprovides a useful summary of variables including numeric and non-numeric datadescribeinpsychprovides descriptive statistics for numeric data
R Example
> library(MASS) # provides dataset called "survey"
> library(Hmisc) # Hmisc describe
> library(psych) # psych describe
The following is the output of Hmisc describe:
> Hmisc::describe(survey)
survey
12 Variables 237 Observations
----------------------------------------------------------------------------------------------------------------------
Sex
n missing unique
236 1 2
Female (118, 50%), Male (118, 50%)
----------------------------------------------------------------------------------------------------------------------
Wr.Hnd
n missing unique Mean .05 .10 .25 .50 .75 .90 .95
236 1 60 18.67 16.00 16.50 17.50 18.50 19.80 21.15 22.05
lowest : 13.0 14.0 15.0 15.4 15.5, highest: 22.5 22.8 23.0 23.1 23.2
----------------------------------------------------------------------------------------------------------------------
NW.Hnd
n missing unique Mean .05 .10 .25 .50 .75 .90 .95
236 1 68 18.58 15.50 16.30 17.50 18.50 19.72 21.00 22.22
lowest : 12.5 13.0 13.3 13.5 15.0, highest: 22.7 23.0 23.2 23.3 23.5
----------------------------------------------------------------------------------------------------------------------
[ABBREVIATED OUTPUT]
Then below is the output of psych describe for the numeric variables:
> psych::describe(survey[,sapply(survey, class) %in% c("numeric", "integer") ])
var n mean sd median trimmed mad min max range skew kurtosis se
Wr.Hnd 1 236 18.67 1.88 18.50 18.61 1.48 13.00 23.2 10.20 0.18 0.36 0.12
NW.Hnd 2 236 18.58 1.97 18.50 18.55 1.63 12.50 23.5 11.00 0.02 0.51 0.13
Pulse 3 192 74.15 11.69 72.50 74.02 11.12 35.00 104.0 69.00 -0.02 0.41 0.84
Height 4 209 172.38 9.85 171.00 172.19 10.08 150.00 200.0 50.00 0.22 -0.39 0.68
Age 5 237 20.37 6.47 18.58 18.99 1.61 16.75 73.0 56.25 5.16 34.53 0.42
answered Aug 5 '11 at 1:20
$\begingroup$
$\endgroup$
I use the codebook function from {EPICALC} which gives summary statistics for a numeric variable and a frequency table with level labels and codes for factors. http://cran.r-project.org/doc/contrib/Epicalc_Book.pdf (see p.50) Moreover, this is very useful because it provides sd for quantitative variables.
Enjoy !
-
1$\begingroup$ +1 (from earlier). I really like the way
codebook()lays this out. 1 issue is thatnas are dropped, which you may want included in your output. 1 way to deal w/ this (at least w/ factors) is to use ?recode.is.na 1st (eg, to "missing"); for numeric variables, you can create a new variable immediately to the left of the column w/ a logical value based onis.na(), then runcodebook(). It is a bit of a kluge, though. $\endgroup$ – gung - Reinstate Monica♦ Dec 4 '12 at 15:35
$\begingroup$
$\endgroup$
You can check out my summarytools package (CRAN link) which includes a codebook-like function, with markdown and html formatting options.
install.packages("summarytools")
library(summarytools)
dfSummary(CO2, style = "grid", plain.ascii = TRUE)
Dataframe Summary
CO2
+------------+---------------+-------------------------------------+--------------------+-----------+
| Variable | Properties | Stats / Values | Freqs, % Valid | N Valid |
+============+===============+=====================================+====================+===========+
| Plant | type:integer | 1. Qn1 | 1: 7 (8.3%) | 84/84 |
| | class:ordered | 2. Qn2 | 2: 7 (8.3%) | (100.0%) |
| | + factor | 3. Qn3 | 3: 7 (8.3%) | |
| | | 4. Qc1 | 4: 7 (8.3%) | |
| | | 5. Qc3 | 5: 7 (8.3%) | |
| | | 6. Qc2 | 6: 7 (8.3%) | |
| | | 7. Mn3 | 7: 7 (8.3%) | |
| | | 8. Mn2 | 8: 7 (8.3%) | |
| | | 9. Mn1 | 9: 7 (8.3%) | |
| | | 10. Mc2 | 10: 7 (8.3%) | |
| | | ... 2 other levels | others: 14 (16.7%) | |
+------------+---------------+-------------------------------------+--------------------+-----------+
| Type | type:integer | 1. Quebec | 1: 42 (50%) | 84/84 |
| | class:factor | 2. Mississippi | 2: 42 (50%) | (100.0%) |
+------------+---------------+-------------------------------------+--------------------+-----------+
| Treatment | type:integer | 1. nonchilled | 1: 42 (50%) | 84/84 |
| | class:factor | 2. chilled | 2: 42 (50%) | (100.0%) |
+------------+---------------+-------------------------------------+--------------------+-----------+
| conc | type:double | mean (sd) = 435 (295.92) | 95: 12 (14.3%) | 84/84 |
| | class:numeric | min < med < max = 95 < 350 < 1000 | 175: 12 (14.3%) | (100.0%) |
| | | IQR (CV) = 500 (0.68) | 250: 12 (14.3%) | |
| | | | 350: 12 (14.3%) | |
| | | | 500: 12 (14.3%) | |
| | | | 675: 12 (14.3%) | |
| | | | 1000: 12 (14.3%) | |
+------------+---------------+-------------------------------------+--------------------+-----------+
| uptake | type:double | mean (sd) = 27.21 (10.81) | 76 distinct values | 84/84 |
| | class:numeric | min < med < max = 7.7 < 28.3 < 45.5 | | (100.0%) |
| | | IQR (CV) = 19.23 (0.4) | | |
+------------+---------------+-------------------------------------+--------------------+-----------+
EDIT
In newer versions of summarytools, the freq() function (which produces straightforward frequency tables, more to-the-point as regards to the original question) accepts data frames as well as single variables. For cross-tabulations (which proc freq also does), see the ctable() function.
freq(CO2)
Frequencies
CO2$PlantType: Ordered Factor
Freq % Valid % Valid Cum % Total % Total Cum
Qn1 7 8.33 8.33 8.33 8.33
Qn2 7 8.33 16.67 8.33 16.67
Qn3 7 8.33 25.00 8.33 25.00
Qc1 7 8.33 33.33 8.33 33.33
Qc3 7 8.33 41.67 8.33 41.67
Qc2 7 8.33 50.00 8.33 50.00
Mn3 7 8.33 58.33 8.33 58.33
Mn2 7 8.33 66.67 8.33 66.67
Mn1 7 8.33 75.00 8.33 75.00
Mc2 7 8.33 83.33 8.33 83.33
Mc3 7 8.33 91.67 8.33 91.67
Mc1 7 8.33 100.00 8.33 100.00
<NA> 0 0.00 100.00
Total 84 100.00 100.00 100.00 100.00
Type: Factor
Freq % Valid % Valid Cum % Total % Total Cum
Quebec 42 50.00 50.00 50.00 50.00
Mississippi 42 50.00 100.00 50.00 100.00
<NA> 0 0.00 100.00
Total 84 100.00 100.00 100.00 100.00
Type: Factor
Freq % Valid % Valid Cum % Total % Total Cum
nonchilled 42 50.00 50.00 50.00 50.00
chilled 42 50.00 100.00 50.00 100.00
<NA> 0 0.00 100.00
Total 84 100.00 100.00 100.00 100.00
answered Nov 9 '14 at 7:47
$\begingroup$
$\endgroup$
Thanks for all the suggestions everyone. I ended up using either table or Rcmdr's numSummary function plus apply:
apply(dataframe[,c('need_rbcs','need_platelets','need_ffp')],2,table)
This works pretty well and is not too inconvenient. However I will definitely give some of these other solutions a try!
