CODEWITHSUNDEEP
rregressionlinear-regressionpredict
jkl
jkl

Reputation: 87

Forecast with regression model with scenarios

I am running AR on this data. 

   Date        Price     YOY    Quarter
   2000-01-15  2.385368 -312362      Q1
   2000-02-15  2.614250 -442117      Q1
   2000-03-15  2.828261 -252596      Q1
   2000-04-15  3.028842 -292756      Q2
   2000-05-15  3.596409 -401578      Q2
   ......

The model is 

price.fit<-lm(log(Price)~+(YOY)+log(lag(Price))+relevel(Quarter,ref="Q4"),subset(Data,Year>=2000))

Now I need to forecast for the next 12 periods. A made-up data frame "newdata" includes estimate(a range) for the variable YOY. 

newdata
high mean low
.....

How to use predict function or other method to forecast the price range(high, mean, low)for the next 12 months? Or maybe using loops?

Upvotes: 3

Views: 994

Answers (1)

Jaehyeon Kim
Jaehyeon Kim

Reputation: 1417

There is a generic function predict() and it can be used for prediction. Although predict.lm() is used below for clarification, predict() should work. Given a lm object and new data, it returns forecasted values.

As regression returns numeric values, converting into a factor (high, medium, low) has to be mapped manually and sapply() is used for that.

library(lubridate)
dates <- ymd(19990115) + months(0:71)
df <- data.frame(date = dates,
                 price = sample(100:150, 72, replace = T),
                 YOY = sample(-40000:-20000, 72, replace = T),
                 Qtr = as.factor(quarter(dates)))

price.fit <- lm(log(price) ~ +(YOY) + log(lag(price)) + relevel(Qtr, ref = "4"),
                data = df[year(df$date) >= 2000,])


pred <- predict.lm(price.fit, newdata = df[year(df$date) < 2000,])
pred
#1        2        3        4        5        6 
#4.976734 4.653960 4.605170 4.691348 4.663439 4.969813 
#7        8        9       10       11       12 
#4.927254 4.727388 4.634729 4.875197 4.700480 4.700480 

sapply(pred, function(x) {
  if(x < 4.56) "low"
  else if(x < 4.75) "medium"
  else "high"
})
#1        2        3        4        5        6        7        8 
#"high" "medium" "medium" "medium" "medium"   "high"   "high" "medium" 
#9       10       11       12 
#"medium"   "high" "medium" "medium"

Upvotes: 1

Related Questions