I have a 3 level contingency table for which I'm trying to calculate the percentages for each cell of the table as a function of the sum row total for each row and then a function of the sum column total for each column. Here is the data which I want to calculate the percentages for:
"A2"
0_15m 15_30m 30_>40m
<35yrs 0_4cm 217 30 3
20_80cm 282 42 14
4_20cm 315 182 82
>=35yrs 0_4cm 334 63 3
20_80cm 310 75 23
4_20cm 433 110 95
dput(A2)
structure(c(217L, 282L, 315L, 334L, 310L, 433L, 30L, 42L, 182L,
63L, 75L, 110L, 3L, 14L, 82L, 3L, 23L, 95L), .Dim = c(6L, 3L), class = "ftable", row.vars = structure(list(
c("<35yrs", ">=35yrs"), c("0_4cm", "20_80cm", "4_20cm")), .Names = c("",
"")), col.vars = structure(list(c("0_15m", "15_30m", "30_>40m"
)), .Names = ""))
I've tried using colPercent function as well as calculating manually (see example):
Where A2 is the ftable of data above:
rpc <- A2 / rowSums(A2) * 100
cpc <- A2 / colSums(A2) * 100
As you can see, the row percentages are calculated correctly (All sum to 100 across the rows), however column percentages are in some cases over 100% and therefore must not have been calculated correctly.
"Row percentages"
0_15m 15_30m 30_>40m
<35yrs 0_4cm 86.800000 12.000000 1.200000
20_80cm 83.431953 12.426036 4.142012
4_20cm 54.404145 31.433506 14.162349
>=35yrs 0_4cm 83.500000 15.750000 0.750000
20_80cm 75.980392 18.382353 5.637255
4_20cm 67.868339 17.241379 14.890282
"Column Percentages"
0_15m 15_30m 30_>40m
<35yrs 0_4cm 11.4754098 1.5864622 0.1586462
20_80cm 56.1752988 8.3665339 2.7888446
4_20cm 143.1818182 82.7272727 37.2727273
>=35yrs 0_4cm 17.6626124 3.3315706 0.1586462
20_80cm 61.7529880 14.9402390 4.5816733
4_20cm 196.8181818 50.0000000 43.1818182
Division works row-wise and not column wise as needed here. To get colSums to work you can transpose, divide and then transpose again
t(t(as.matrix(A2))/colSums(A2)) * 100
# 0_15m 15_30m 30_>40m
#
#<35yrs 0_4cm 11.48 5.98 1.36
# 20_80cm 14.91 8.37 6.36
# 4_20cm 16.66 36.25 37.27
#>=35yrs 0_4cm 17.66 12.55 1.36
# 20_80cm 16.39 14.94 10.45
# 4_20cm 22.90 21.91 43.18
Or another option is
A2 / colSums(A2)[col(A2)] * 100