如何通过R中的TimeDate组计算条件平均值（音量加权）？

Question

I am trying to calculate volume weighted average price (VWAP) from a trade data, based on same DateTime group. The sample data from a small data frame (20,000 entries) is as following:

testdata[,c(5,8,10,11)]
               transactiontime price volume totalEquity
334 2014-12-01 01:30:00.110000 19330 500000      966500
335 2014-12-01 01:30:00.830000 19340   8000       15472
336 2014-12-01 01:30:00.830000 19340   1000        1934
337 2014-12-01 01:30:00.830000 19340   1000        1934
338        2014-12-01 01:30:01 19340    500         967
339        2014-12-01 01:30:01 19340   2000        3868
340        2014-12-01 01:30:01 19340   4000        7736
341        2014-12-01 01:30:01 19340  40000       77360
342        2014-12-01 01:30:01 19340 500000      967000
343        2014-12-01 01:30:01 19340  12000       23208

where totalEquity was an added column by workdata$totalEquity <- (workdata$price)/10000 * (workdata$volume)

I want to calculate VWAP by transaction time. Using aggregate is easy to calculate mean(totalEquity) , but how can I calculate the volume weighted average so to achieve something like:

group.1            transactiontime weightedPrice
1       2014-12-01 01:30:00.110000         1.933
2       2014-12-01 01:30:00.830000         1.934
3              2014-12-01 01:30:01         1.934

where weightedPrice is sum(totalEquity)/sum(volume) , grouped by transactiontime .

I've searched for many questions on group mean but did not know how to combine each function correctly; none of my previous attempts worked, as of following, and I was very frustrated:

volWeighted <- function(x=workdata$totalEquity,y=workdata$volume) {sum(x)/sum(y)}
aggregate(totalEquity~transactiontime, testdata, FUN=volWeighted)

or

library(data.table)
dt[,list(avg_tte <- sum(testdata$totalEquity)/sum(testdata$volume)),
'testdata$transactiontime']

or

setDT(testdata) [, time.diff :=max(time)-min(time), by=transactiontime]
                [, if(time.diff==0)
                      .( totalEquity = sum(totalEquity)/sum(volume))
                    else .SD, by = .(transactiontime, time.diff)]

This is my first question and I tried to make it efficient, but if it happens to be a duplicate, please let me know and I am very willing to learn from the previous question.

Answer 1

Close on all your attempts.

Instead of aggregate try by :

by(workdata, workdata$transactiontime, function (x) sum(x$totalEquity)/sum(x$volume))

This will return just the values in of the weightedPrice column - up to you to add it to the dataframe.

I'd recommend something like plyr or data.frame

library(plyr)
ddply(workdata, .(transactiontime), summarize,
      weightedPrice=sum(totalEquity)/sum(volume))

The summarize function calculates some summary statistic, and ddply calls summarize for each unique transactiontime . The calculation works a bit like the transform function, where if you write weightedPrice=sum(totalEquity)/sum(volume) it calculates the expression looking up columns in workdata (so no need for workdata$columnname ), and assigns it to a column called weightedPrice .

If data.table :

library(data.table)
setDT(workdata) # turn workdata into data.table
workdata[, list(weightedPrice=sum(totalEquity)/sum(volume)), by=transactiontime]