CODEWITHSUNDEEP
r
user37352
user37352

Reputation:

How to calculate Z-score by group

I want to calculate Z-scores using means and standard deviations generated from each group. For example I have following table. It has 3 groups of data, I can generate mean and standard deviation for each group. Then I use group 1 mean and SD to calculate the Zscores for group one data points, and etc...

> dat
   group level    y
1      1     A 10.8
2      1     B 12.0
3      1     C  9.6
4      1     A 12.0
5      1     B  7.8
6      1     C 10.8
7      2     A  8.7
8      2     B  9.2
9      2     C  8.2
10     2     A 10.0
11     2     B 12.2
12     2     C  8.2
13     3     A 10.9
14     3     B  8.3
15     3     C 10.1
16     3     A  9.9
17     3     B 10.9
18     3     C 10.3

I have learned from this blog on how to get summary data by group, but not sure how to go from there.

Thanks.

Upvotes: 9

Views: 16293

Answers (4)

dmt
dmt

Reputation: 2183

In dplyr

library(dplyr)

dat_z = dat %>%
        group_by(group) %>%
        mutate(z_score = scale(y))

Upvotes: 3

ndoogan
ndoogan

Reputation: 1925

Base R (i.e., no dependencies required) includes the functions ave() (for group wise application) and scale() (for calculating z-scores):

dat$z <- ave(dat$y, dat$group, FUN=scale)

Then the new variable z in dat will contain the groupwise-scaled variable.

Note that unlike similar functions in Base R (e.g., sapply, lapply), you need to include FUN= explicitly.

Upvotes: 14

RHelp
RHelp

Reputation: 835

You can use the ddply function of plyr and calculate the z score.

library(plyr)
dat <-  ddply(dat, .(group), summarize, z_score=scale(y))

or you can calculate it manually as -

dat <-  ddply(dat, .(group), summarize, z_score=(y-mean(y))/sd(y)))

If you have na's in your data, then add na.rm=True in the mean and sd functions.

Hope this helps.

Upvotes: 3

dmartin
dmartin

Reputation: 653

I would check out data.table for this.

Something like:

require(data.table)
datDT <- data.table(dat)
datDT[, yScaled := scale(y), by = group]

Upvotes: 6

Related Questions