0

I have this ts: 

   Point       Forecast      Lo 80    Hi 80      Lo 95    Hi 95
2015.173       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.192       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.212       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.231       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.250       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.269       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.288       1.657143 0.06698996 3.247296 -0.7747861 4.089072
2015.308       1.657143 0.06698996 3.247296 -0.7747861 4.089072

How can I convert it in ts, I am using:

fct <- data.frame(date=as.Date(index(fca)), Y = melt(fca))

But I get this error:

Error in data.frame(date = as.Date(index(fca)), Y = melt(fca)) :
arguments imply differing number of rows: 10, 16

Someone know how to fix it?

Psidom
  • 209,562
  • 33
  • 339
  • 356
Erika
  • 1

1 Answers1

0

The forecast output only looks like a data frame when you print it. But if you look at the structure with str(fca), you will see the messy guts of the object.

What you see when you call print(fca) is a specially tailored output as a visual aid ONLY. It is not meant to hint that the underlying object is a matrix-like array.

The solution is to extract the parts you are after and create a data frame from them.

library(forecast)
fit <- Arima(WWWusage,c(3,1,0))
fca <- forecast(fit)
cbind.data.frame(Point=fca$mean, Lower=fca$lower, Upper=fca$upper)
#           Point Lower.80% Lower.95% Upper.80% Upper.95%
# 1   0.005151116 -1.271415 -1.947189  1.281718  1.957491
# 2  -0.170997839 -1.716453 -2.534568  1.374457  2.192572
# 3  -0.205288123 -1.782216 -2.616991  1.371640  2.206415
# 4  -0.221099340 -1.812749 -2.655317  1.370550  2.213119
# 5  -0.229741679 -1.830356 -2.677670  1.370872  2.218186
# 6  -0.234764205 -1.841557 -2.692142  1.372029  2.222614
# 7  -0.237705586 -1.849083 -2.702094  1.373671  2.226683
# 8  -0.239354235 -1.854302 -2.709204  1.375594  2.230495
# 9  -0.240158452 -1.857987 -2.714414  1.377670  2.234097
# 10 -0.240391964 -1.860607 -2.718298  1.379824  2.237514

Edit

The dates are hidden pretty well in the data. I did not find an exact date vector. I believe it is generated by the tsp attribute indicating the start, end, and length:

dates <- attr(fca$mean, "tsp")
datecol <- seq(from=dates[1], to=dates[2], by=dates[3])
Pierre L
  • 28,203
  • 6
  • 47
  • 69