I want to create a ts() object from a dataframe for forecasting physical phenomena.
My data has a 30 min frequency over a period of 1 year (1-1-2018 to 12-31-2018) also I have observed that my data has a seasonality of 1 day.
> head(pleiadesGH.v2[,c("time", "humExt.R", "tempExt", "radExt", "vientoVelo")])
time humExt.R tempExt radExt vientoVelo
1 2018-01-01 00:00:00 NA NA NA NA
2 2018-01-01 00:30:00 36.78287 16.95125 -10.08125 3.68550
3 2018-01-01 01:00:00 38.56775 16.26350 -9.75000 2.38420
4 2018-01-01 01:30:00 38.76425 15.63470 -10.08125 2.71915
5 2018-01-01 02:00:00 39.61575 15.32030 -10.41250 3.70475
6 2018-01-01 02:30:00 37.48700 15.06485 -10.74375 2.51895
Based on this answers :
https://robjhyndman.com/hyndsight/seasonal-periods/
time series with 10 min frequency in R
I conclude that my ts() frequency should be of 48, because of 1 day have 48 observations.
ts.freq1 <- ts(data = pleiadesGH.v2[,2:ncol(pleiadesGH.v2)],
start = c(2018),
frequency = 48)
But the resulting ts() has a wrong time index as you can see below. The time data should be from 2018 to 2019 instead of 2400.
Time Series:
Start = c(2018, 1)
End = c(2383, 1)
Frequency = 48
humInt.R humInt.E tempInt tempMac humExt.R humExt.E radExt tempExt vientoVelo
2018.000 NA NA NA NA NA NA NA NA NA
2018.021 NA NA NA NA 36.78287 0.004410894 -10.08125 16.95125 3.6855000
2018.042 NA NA NA NA 38.56775 0.004427114 -9.75000 16.26350 2.3842000
2018.062 NA NA NA NA 38.76425 0.004273306 -10.08125 15.63470 2.7191500
2018.083 NA NA NA NA 39.61575 0.004280005 -10.41250 15.32030 3.7047500
2018.104 NA NA NA NA 37.48700 0.003982139 -10.74375 15.06485 2.5189500
2018.125 NA NA NA NA 35.84950 0.003735063 -10.41250 14.77010 3.2235000
2018.146 NA NA NA NA 36.68462 0.003697674 -8.75625 14.25920 1.4409500
2018.167 NA NA NA NA 41.48250 0.003954404 -11.07500 13.39460 1.5064000
2018.188 NA NA NA NA 42.54688 0.003968433 -9.41875 13.06055 3.6701000
2018.208 NA NA NA NA 43.05450 0.003969581 -9.08750 12.88370 1.6103500
2018.229 NA NA NA NA 44.11888 0.004000366 -9.41875 12.62825 1.3485500
2018.250 NA NA NA NA 46.26400 0.004061953 -9.08750 12.13700 1.9491500
2018.271 NA NA NA NA 46.88625 0.004084874 -9.08750 12.01910 2.0569500
2018.292 NA NA NA NA 49.57175 0.004187059
I also tried with this frequency:
ts.freq1 <- ts(data = pleiadesGH.v2[,2:ncol(pleiadesGH.v2)],
start = c(2018),
frequency = 365.25*24*60/30 )
Getting the following result :
Time Series:
Start = c(2018, 1)
End = c(2018, 17521)
Frequency = 17532
humInt.R humInt.E tempInt tempMac humExt.R humExt.E radExt tempExt vientoVelo
2018.000 NA NA NA NA NA NA NA NA NA
2018.000 NA NA NA NA 36.78287 0.004410894 -10.08125 16.95125 3.6855000
2018.000 NA NA NA NA 38.56775 0.004427114 -9.75000 16.26350 2.3842000
2018.000 NA NA NA NA 38.76425 0.004273306 -10.08125 15.63470 2.7191500
2018.000 NA NA NA NA 39.61575 0.004280005 -10.41250 15.32030 3.7047500
2018.000 NA NA NA NA 37.48700 0.003982139 -10.74375 15.06485 2.5189500
2018.000 NA NA NA NA 35.84950 0.003735063 -10.41250 14.77010 3.2235000
2018.000 NA NA NA NA 36.68462 0.003697674 -8.75625 14.25920 1.4409500
2018.000 NA NA NA NA 41.48250 0.003954404 -11.07500 13.39460 1.5064000
2018.001 NA NA NA NA 42.54688 0.003968433 -9.41875 13.06055 3.6701000
2018.001 NA NA NA NA 43.05450 0.003969581 -9.08750 12.88370 1.6103500
2018.001 NA NA NA NA 44.11888 0.004000366 -9.41875 12.62825 1.3485500
2018.001 NA NA NA NA 46.26400 0.004061953 -9.08750 12.13700 1.9491500
But this implicitly means that my seasonality is yearly but this is not my objective. In the following picture you can se the time index is now fixed despite the wrong seasonality
good index incorrect seasonality
What im doing wrong?
The solution is as follows:
freq.daily <- 48 # 24 hours * 2 obs per hour
ts.daily <- ts(data = pleiadesGH.v2.interp[,2:ncol(pleiadesGH.v2)],
start = c(1),
frequency = freq.daily)
Time Series:
Start = c(1, 1)
End = c(366, 1)
Frequency = 48
humInt.R humInt.E tempInt tempMac humExt.R humExt.E radExt
1.000000 74.56250 0.007699896 14.53500 13.625000 36.78287 0.004410894 -10.081250
1.020833 74.56250 0.007699896 14.53500 13.625000 36.78287 0.004410894 -10.081250
1.041667 74.56250 0.007699896 14.53500 13.625000 38.56775 0.004427114 -9.750000
1.062500 74.56250 0.007699896 14.53500 13.625000 38.76425 0.004273306 -10.081250
1.083333 74.56250 0.007699896 14.53500 13.625000 39.61575 0.004280005 -10.412500
1.104167 74.56250 0.007699896 14.53500 13.625000 37.48700 0.003982139 -10.743750
As this is the way ts simply and effectively manage dates, starting from 1.