如何估算Y轴的值，以便我在R中的时间序列（X轴）上具有相等的间隔

Question

I have a table in MYSQL with datetime object and contains Columns (Price) which is specific price at that particular time.

[1] "2015-06-25 10:33:02 IST" "2015-06-25 10:35:32 IST"
[3] "2015-06-25 10:38:02 IST" "2015-06-25 10:40:02 IST"
[5] "2015-06-25 10:42:02 IST" "2015-06-25 10:44:02 IST"
[7] "2015-06-25 10:46:02 IST" "2015-06-25 10:49:02 IST"
[9] "2015-06-25 10:51:32 IST" "2015-06-25 10:53:02 IST"
[11] "2015-06-25 10:55:32 IST" "2015-06-25 10:57:32 IST"
[13] "2015-06-25 10:59:32 IST" "2015-06-25 11:01:02 IST"
[15] "2015-06-25 11:03:32 IST" "2015-06-25 11:05:32 IST"
[17] "2015-06-25 11:07:32 IST" "2015-06-25 11:09:02 IST"

Now as you see the time interval is not constant for each row in table. I want to plot the graph between Price and time, I want to keep the X-axis(Time) in fixed intervals and will try to plot estimated Y value(Price)

For Ex:
Time series: 10:33:00, 10:35:00, 10:37:00, 10:39:00 etc
Price(Estimated): ??

I need help on following:

1. Howto get Estimated value of Y-axis considering two nearby price.
2. I want to plot all these fixed interval points but X-Axis should be displaying scale in interval of 2 hours.

Below is my code, I am very new to R so please excuse if I am asking something very basic.

x<-data[,"idUnique"]
y<-data[,"Price"]
z<-data[,"datetime"]
z<-strptime(z, format="%d-%b-%Y %H:%M:%S",tz="IST") 
meanpivot<-mean(y)
meanx<-mean(x)
plot(x, y, type="l", col="red", xlim=c(meanx-400, meanx+400), ylim=c(meanpivot*0.99, meanpivot*1.01))

Right now I am using IDUnique, for X axis but I want to use Time(z) going forward.

Any help is appreciated. Thanks in advance !!

Answer 1

As a side note for future questions - it would be lovely to have a small reproducible example of your problem. We can't copy-paste your code so far to try it out, because we don't have access to data . My answer demonstrates how you could have made a reproducible example (ie you can copy-paste it directly into your R console and it will work).

The gist of the answer is to use approx to interpolate your per-interval prices, and then plot them using the package of your choice, adjusting the x axis intervals to whatever yo uwant.

As far as I understand your question, your data has a column of datetimes and a column of prices, eg:

# make an example dataframe with one value per (approximately) minute
x <- data.frame(datetime = as.POSIXct('2015-07-15 13:48:42') + jitter((1:60)*60),
                price = runif(60))

Now I want to calculate the price every 2 minutes (according to your example). First we generate a vector of times we want to calculate at:

# a vector of times from 13:50 every 2 minutes for an hour.
times.2min <- as.POSIXct('2015-07-15 13:50:00') + (0:29)*2*60

Note - you can't interpolate for times outside the raneg of your original datetimes, so be careful of that when you create your times.2min -equivalent.

Now to calculate the price at these times, we can do an interpolation. I'll just do linear approximation; see ?approx for further details.

prices <- approx(x$datetime, x$price, xout=times.2min)

prices$x has our times.2min and prices$y has the calculated prices for those times.

Then it's just a matter of plotting:

plot(y ~ x, prices, type='l')

The x axis ticks are automatically determined to not be too cluttered. If you want to specify them yourself (eg I'll do every 20 minutes here)

x.axis.ticks <- seq(from=round(min(prices$x), 'hour'),
                    to=round(max(prices$x), 'hour'),
                    by=20*60) # 14:00, 14:20, 14:40, 15:00
# plot points, skip x axis
plot(y ~ x, prices, type='l', xaxt='n')
# plot x axis
axis(1, at=x.axis.ticks, labels=as.character(x.axis.ticks))

Or if you like use ggplot to get something prettier

library(ggplot2)
ggplot(data.frame(prices), aes(x=x, y=y)) +
    geom_line() +
    scale_x_datetime(breaks=date_breaks(width="20 min"))