从 R 中的每日温度数据创建时间序列

Question

I have a 24 year data set (from 1/1/1980 to 31/12/2014) of daily temperature data in a data frame. The form of my data frame is as follows:

date	st1
1980-1-1	-2.3
1980-1-2	-1.2
1980-1-3	0.8
...	...
2014-12-31	4.7

( "..." stands for sequence of data from 1980 - 2014) In order to use the time series packages I need to convert the data frame in time series class. I used the following code:

temp_ts <- ts(temp_data$st1,start = c(1980,1,1), end = c(2014,12,31),frequency = 12)

I get a complete time series, with all months in the heading of each column, and a year in each row in form of a table, as follows:

	Jan	Feb	Mar	Apr	May	June	..
1980	1.8	-1.0		-4.8	-5.2	-3.1	..
1981	-5.0	-5.2	-3.8	-0.5	0.3	0.2	..
...	...	...	...	...	...	...	...
2014	5.3	6.3	-1.7	-4.3	-6.7	-4.7	..

( there is no "..." in my table, i just use it here to save some space)

The problem is that the data in this time series are wrong. More specifically, in May the temperatures in my original data are >20 C, so it can't have as an average value negative values. Same thing applies to all other months. This means that in my time series, each month column doesn't have the appropriate average value of each month but something else which is I'm not capable to identify how it was calculated.

I would appreciate every help in order to identify and solve this mistake by creating a time series in which each value corresponds to the average value of corresponding year and month.

Answer 1

Convert the data to monthly data first. Then transform the monthly data into a ts. Try the following command according to convert the data to monthly data:

df %>%
group_by(month = floor_date(date, "month")) %>%
summarize(temp = mean(temp))

Save this df into a variable and then transform that into a ts. I hope this might work for you.