Creating Time Series grouped by variable levels - time-series

Question

I'm trying to create a time series that will predict scores grouped by team.
TeamScores$Year <- as.Date(TeamScores$Year)
sample <-TeamScores[1:20,]
dput(sample)
structure(list(
Team = c("Abl Christian", "Air Force", "Akron", "Alab A&M", "Alabama", "Alabama St", "Albany", "Alcorn State", "American", "App State", "AR Lit Rock", "Arizona", "Arizona St", "Ark Pine Bl", "Arkansas", "Arkansas St", "Army", "Auburn", "Austin Peay", "Ball State"),
Score = c(71.7, 67.4, 68.4, 60.6, 71.8, 65.6, 66.8, 60.3, 72, 77.3, 73.6, 70.9, 77.8, 65.3, 75.5, 72.8, 70.2, 78.9, 80.1, 74.1),
Year = structure(
c(17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532, 17532),
class = "Date")),
row.names = c(NA, -20L),
class = c("tbl_df", "tbl", "data.frame"))
I created a time series successfully (I think), but I cannot get my fit to work.
time_ser<-ts(matrix(TeamScores$Team,nrow=3530),start=c(2009-01-01),frequency=1)
class(time_ser)
#[1] "ts"
fit<- auto.arima(time_ser)
#Error in stats::arima(x = x, order = order, seasonal = seasonal, include.mean = include.mean, :
'x' must be numeric
In addition: Warning message:
In is.constant(x) : NAs introduced by coercion
My x (Score) is numeric and I'm just lost. I assumed that I needed to do the auto.arima function to then perform the predict function.

Answer 1

With the datastructure of your format you can run ARIMA like this:
# making the structure
TeamScores <- structure(list(
Team = c("Abl Christian", "Air Force", "Akron", "Abl Christian", "Air Force", "Akron","Abl Christian", "Air Force", "Akron","Abl Christian", "Air Force", "Akron","Abl Christian", "Air Force", "Akron","Abl Christian", "Air Force", "Akron","Abl Christian", "Air Force"),
Score = c(71.7, 67.4, 68.4, 60.6, 71.8, 65.6, 66.8, 60.3, 72, 77.3, 73.6, 70.9, 77.8, 65.3, 75.5, 72.8, 70.2, 78.9, 80.1, 74.1),
Year = structure(
c(17532, 17533, 17534, 17535, 17536, 17537, 17538, 17539, 17540, 17541, 17542, 17543, 17544, 17545, 17546, 17547, 17548, 17549, 17550, 17551),
class = "Date")),
row.names = c(NA, -20L),
class = c("tbl_df", "tbl", "data.frame"))
# make a vector with team names:
teamnames <- c("Abl Christian", "Air Force", "Akron")
# run ARIMA for each team:
for (team in teamnames){
subdf <- subset(TeamScores, Team==team)
fit <- auto.arima(subdf$Score,xreg=subdf$Year)
print(fit)}
P.S. I couldn´t run arima with your sample code/data, because in the sample all of the dates are the same (2018-01-01) and each group is present only once, and you cannot really make a time series with one timepoint nor one datapoint per group. Thus I´ve made a similar structure for the testing. Also I skipped making the ts object and ran ARIMA directly on the dataframe.