R 从数据帧中选择范围（时间戳），然后再进行总结

使用来自的信息，我可以使用时间戳信息总结我的数据帧

data <- read.csv("data.csv", header=T)
data$TIMESTAMP <- strptime(data$TIMESTAMP, "%m/%d/%Y %H:%M") # make unambigous
data$TIMESTAMP <- as.POSIXct(data$TIMESTAMP) # plyr/ddply does not seem to like POSIXlt
data$DAY <- as.factor(format(data$TIMESTAMP,'%d'))
data$MONTH <- as.factor(format(data$TIMESTAMP,'%m'))
ddply(data,.(MONTH,DAY),summarise, V1 = mean(P), V2 = max(WS)) # summarize by month by day

我在evalexpr、envir、enclose中得到以下错误：找不到对象“MONTH”。我能看到的data和data.1.4之间的唯一区别是row.names是“可见”的非顺序数据。1.4 dataframe

那么，如何从数据帧中子集或选择条目来创建一个新的df，然后进行汇总

# dput of subset of data
data <- structure(list(TIMESTAMP = structure(1:15, .Label = c("1/1/2012 11:00", 
"1/1/2012 12:00", "1/1/2012 13:00", "1/1/2012 14:00", "1/1/2012 15:00", 
"1/2/2012 11:00", "1/2/2012 12:00", "1/2/2012 13:00", "1/2/2012 14:00", 
"1/2/2012 15:00", "4/7/2012 11:00", "4/7/2012 12:00", "4/7/2012 13:00", 
"4/7/2012 14:00", "4/7/2012 15:00"), class = "factor"), P = c(992.4, 
992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 
239, 239, 239, 239, 239), WS = c(4.023, 3.576, 4.023, 6.259, 
4.47, 3.576, 3.576, 2.682, 4.023, 3.576, 2.682, 3.129, 2.682, 
2.235, 2.682), WD = c(212L, 200L, 215L, 213L, 204L, 304L, 276L, 
273L, 307L, 270L, 54L, 24L, 304L, 320L, 321L), AT = c(16.11, 
18.89, 20, 20, 19.44, 10.56, 11.11, 11.67, 12.22, 11.11, 17.22, 
18.33, 19.44, 20.56, 21.11), FT = c(17.22, 22.22, 22.78, 22.78, 
20, 11.11, 15.56, 17.22, 17.78, 15.56, 24.44, 25.56, 29.44, 30.56, 
29.44), H = c(50L, 38L, 38L, 39L, 48L, 24L, 19L, 18L, 16L, 18L, 
23L, 20L, 18L, 17L, 15L), B = c(1029L, 1027L, 1026L, 1024L, 1023L, 
1026L, 1025L, 1024L, 1023L, 1023L, 1034L, 1033L, 1032L, 1031L, 
1030L), FM = c(14.9, 14.4, 14, 13.7, 13.6, 13.1, 12.8, 12.3, 
12, 11.7, 12.8, 12, 11.4, 10.9, 10.4), GD = c(204L, 220L, 227L, 
222L, 216L, 338L, 311L, 326L, 310L, 273L, 62L, 13L, 312L, 272L, 
281L), MG = c(8.047, 9.835, 10.28, 13.41, 11.18, 9.388, 8.941, 
8.494, 9.835, 10.73, 6.706, 7.153, 8.047, 8.047, 7.6), SR = c(522L, 
603L, 604L, 526L, 248L, 569L, 653L, 671L, 616L, 487L, 972L, 1053L, 
1061L, 1002L, 865L), WS2 = c(2.235, 3.576, 4.47, 4.47, 5.364, 
4.023, 2.682, 3.576, 3.576, 4.023, 3.129, 3.129, 3.576, 2.682, 
3.129), WD2 = c(200L, 201L, 206L, 210L, 211L, 319L, 315L, 311L, 
302L, 290L, 49L, 39L, 15L, 348L, 329L)), .Names = c("TIMESTAMP", 
"P", "WS", "WD", "AT", "FT", "H", "B", "FM", "GD", "MG", "SR", 
"WS2", "WD2"), class = "data.frame", row.names = c(NA, -15L))

---

=13和小时数的所有列取平均值，那么您的示例代码对我来说运行良好，至少当我使用dput输出代替csv中的读取时，代码会更小。没有关于缺少月份对象的错误消息。您在使用数据时没有收到错误消息。1.4数据帧？不，ddplydata.1.4、.MONTH、DAY、summary，V1=meanP，V2=maxWS运行正常。也许可以清除你的工作区，然后再次运行你的代码？@nofunsally，它对我来说也运行得很好。然而，我在另一篇关于POSIXlt的文章中注意到了plyr和您的代码的问题。你能粘贴你的会话信息吗？

# dput of subset of data
data <- structure(list(TIMESTAMP = structure(1:15, .Label = c("1/1/2012 11:00", 
"1/1/2012 12:00", "1/1/2012 13:00", "1/1/2012 14:00", "1/1/2012 15:00", 
"1/2/2012 11:00", "1/2/2012 12:00", "1/2/2012 13:00", "1/2/2012 14:00", 
"1/2/2012 15:00", "4/7/2012 11:00", "4/7/2012 12:00", "4/7/2012 13:00", 
"4/7/2012 14:00", "4/7/2012 15:00"), class = "factor"), P = c(992.4, 
992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 992.4, 
239, 239, 239, 239, 239), WS = c(4.023, 3.576, 4.023, 6.259, 
4.47, 3.576, 3.576, 2.682, 4.023, 3.576, 2.682, 3.129, 2.682, 
2.235, 2.682), WD = c(212L, 200L, 215L, 213L, 204L, 304L, 276L, 
273L, 307L, 270L, 54L, 24L, 304L, 320L, 321L), AT = c(16.11, 
18.89, 20, 20, 19.44, 10.56, 11.11, 11.67, 12.22, 11.11, 17.22, 
18.33, 19.44, 20.56, 21.11), FT = c(17.22, 22.22, 22.78, 22.78, 
20, 11.11, 15.56, 17.22, 17.78, 15.56, 24.44, 25.56, 29.44, 30.56, 
29.44), H = c(50L, 38L, 38L, 39L, 48L, 24L, 19L, 18L, 16L, 18L, 
23L, 20L, 18L, 17L, 15L), B = c(1029L, 1027L, 1026L, 1024L, 1023L, 
1026L, 1025L, 1024L, 1023L, 1023L, 1034L, 1033L, 1032L, 1031L, 
1030L), FM = c(14.9, 14.4, 14, 13.7, 13.6, 13.1, 12.8, 12.3, 
12, 11.7, 12.8, 12, 11.4, 10.9, 10.4), GD = c(204L, 220L, 227L, 
222L, 216L, 338L, 311L, 326L, 310L, 273L, 62L, 13L, 312L, 272L, 
281L), MG = c(8.047, 9.835, 10.28, 13.41, 11.18, 9.388, 8.941, 
8.494, 9.835, 10.73, 6.706, 7.153, 8.047, 8.047, 7.6), SR = c(522L, 
603L, 604L, 526L, 248L, 569L, 653L, 671L, 616L, 487L, 972L, 1053L, 
1061L, 1002L, 865L), WS2 = c(2.235, 3.576, 4.47, 4.47, 5.364, 
4.023, 2.682, 3.576, 3.576, 4.023, 3.129, 3.129, 3.576, 2.682, 
3.129), WD2 = c(200L, 201L, 206L, 210L, 211L, 319L, 315L, 311L, 
302L, 290L, 49L, 39L, 15L, 348L, 329L)), .Names = c("TIMESTAMP", 
"P", "WS", "WD", "AT", "FT", "H", "B", "FM", "GD", "MG", "SR", 
"WS2", "WD2"), class = "data.frame", row.names = c(NA, -15L))

# changing the name from `TIMESTAMP` to `ts` : too long for me!
names(data)[1] <- "ts"

require(data.table)
dt <- data.table(data) # convert to data.table
dt[, ts := as.POSIXct(ts, format="%m/%d/%Y %H:%M")] # convert to POSIXct

# get hour, day, month
dt[, `:=`(day=as.POSIXlt(ts)$mday + 1, 
          month=as.POSIXlt(ts)$mon + 1, 
          hour=as.POSIXlt(ts)$hour)]
# set keys for blazing fast access
setkey(dt, "day","month")

# the `by` part groups your data.table by day and month
# The group can be accessed by `.SD` and we filter there for hours
# then we select all the columns that we want using `with=FALSE` that makes 
# data.table behave like a data.frame only for that operation
# then we use `lapply` to take the mean of all the columns.
# all this can be done in one line! That's the sweetest part!

dt[, lapply(.SD[hour >= 13 & hour <= 15][, 2:13, with=FALSE], mean), by = "day,month"]

#    day month     P       WS       WD       AT       FT        H        B       FM       GD        MG       SR   WS2
# 1:   2     1 992.4 4.917333 210.6667 19.81333 21.85333 41.66667 1024.333 13.76667 221.6667 11.623333 459.3333 4.768
# 2:   3     1 992.4 3.427000 283.3333 11.66667 16.85333 17.33333 1023.333 12.00000 303.0000  9.686333 591.3333 3.725
# 3:   8     4 239.0 2.533000 315.0000 20.37000 29.81333 16.66667 1031.000 10.90000 288.3333  7.898000 976.0000 3.129

dt1 <- dt[, .SD[hour >= 13 & hour <= 15], by="day,month"]
dt1[, lapply(.SD, mean), by="day,month", .SDcols=4:17]