CODEWITHSUNDEEP
rlm
Ojaswita
Ojaswita

Reputation: 93

Predicting using an exponential model

I have the following data:

Days    Total cases
1   3
2   3
3   5
4   6
5   28
6   30
7   31
8   34
9   39
10  48
11  63
12  70
13  82
14  91
15  107
16  112
17  127
18  146
19  171
20  198
21  258
22  334
23  403
24  497
25  571
26  657
27  730
28  883
29  1024
30  1139
31  1329
32  1635
33  2059
34  2545
35  3105
36  3684
37  4289
38  4778
39  5351
40  5916
41  6729
42  7600
43  8452
44  9210
45  10453
46  11484
47  12370
48  13431
49  14353
50  15724
51  17304
52  18543
53  20080
54  21372

I defined days as 'days' and total cases as 'cases1'. I run the following code:

exp.mod <- lm(log(cases1)~days)

I get a good model with reasonable residuals and p-value.

but when i run the following:

predict(exp.mod, data.frame(days=60))

I get the value of 11.66476, which doesnt seem to be correct.

I need to get the value and also include the predictive plot in the exponential model. Hope that clarifies the issue.

Upvotes: 0

Views: 316

Answers (2)

Produnis
Produnis

Reputation: 567

I would do it like this:

# predictr cases from n days
exp(predict(exp.mod, list(days=60)))

to predict days to reach n cases, I would do

exp.mod <- lm(log(cases1) ~ days)
a <- exp(exp.mod$coefficients[1])
b <- exp.mod$coefficients[2]
# Prediction for - lets say - 1500 cases
(log(1500) - log(a)) / b

Upvotes: 0

Earl Mascetti
Earl Mascetti

Reputation: 1336

you should consider the EST models from the forecast package.

Below an example.

library(dplyr)
library(forecast)
ausair %>% ets() %>% forecast() %>% autoplot()

I suggest you to check the free book of the Prof. Rob J Hyndman and Prof George Athanasopoulos wrote (are the authors of the forecast package).

Upvotes: 1

Related Questions