Loops for forecasting

[saila]

Hello, I have panel data for some kidney centers. now I need to do linear forecasting them. How can I not using the repetition of coding and be able to do them in 1 coding set up?

Data in panel format, 20 kidney center+6 periods.

See the code below:
KTV.raw<-read.csv(file="KTV_Reporting(1).csv")

library(stargazer)
library(fpp3)
library(ggplot2)
#Auburn kidney center
Auburn.data<-KTV.raw%>%filter(ClinicName=="Auburn Kidney Center")
ggplot(Auburn.data)+ geom_line(mapping = aes(y=Score, x= Month))+
labs(x="Month", y="Score")+
ggtitle("Auburn Kidney Center KTV Matrix")

Auburn.data1<-as_tsibble(Auburn.data, index=Month)
autoplot(Auburn.data1, Score)

#Auburn.data1|>filter(year(Month) >= 2022)

forecast_Auburn <- Auburn.data1 |>
model(TSLM(Score~trend()))
fc_Auburn <- forecast(forecast_Auburn,h=6)

fc_Auburn |>
autoplot(Auburn.data1,show_gap= FALSE) +
labs(
title = "Forecasts of KTV Matrix-Auburn Kidney Center",
y = "KTV Score")

#Bellevue Kidney center
Bellevue.data<-KTV.raw%>%filter(ClinicName=="Bellevue Kidney Center")
Bellevue.data1<-as_tsibble(Bellevue.data, index=Month)
forecast_Bellevue <- Bellevue.data1 |>
model(TSLM(Score~trend()))

fc_Bellevue <- forecast(forecast_Bellevue,h=6)
fc_Bellevue |>
autoplot(Bellevue.data1,show_gap= FALSE) +
labs(
title = "Forecasts of KTV Matrix-Bellevue Kidney Center",
y = "KTV Score")

[nirgrahamuk]

Here is a code example of how to reduce code repetition, using a reproducible example.
I hope that if you work through this example you will see it shows a general approach to a common problem or reducing repetative code that was likely made by copy-pasting and changing names.

you want to extract Auburn related info , then you want to do the same with Bellevue, etc.
Here is an analogous example using a common data example for R users called iris which has information about 3 species of plant

# lets say I want to do seperate ggplots from iris for each species

library(tidyverse)
# naive attempt 1
 
setosa <- filter(iris,Species=="setosa")
setosa_plot <- ggplot(data=setosa)+aes(x=Petal.Length,
                                       y=Petal.Width)+ geom_point()

versicolor <- filter(iris,Species=="versicolor")
versicolor_plot <- ggplot(data=versicolor)+aes(x=Petal.Length,
                                               y=Petal.Width)+ geom_point()


virginica <- filter(iris,Species=="virginica")
virginica_plot <- ggplot(data=virginica)+aes(x=Petal.Length,
                                             y=Petal.Width)+ geom_point()

#this works ; it makes 3 plots, but is heavily repetitive

# step 1 is to *rewrite as a function*

do_one <- function(x){
  temp<- filter(iris,Species==x) # we parameterise the Species now its whatever x is
  temp_plot<- ggplot(data=temp)+aes(x=Petal.Length,
                                         y=Petal.Width)+ geom_point()
  temp_plot # returning an object that is the result ; its name is not important and can be the same each time
}
# using manually
(set_p <- do_one("setosa"))
(ver_p <- do_one("versicolor"))
(vir_p <- do_one("virginica"))

# This is clearly a step in the right direction, but we are still repeating the calls 3 times, and 
#we may have analyis that requires us to do it 1000 times; so why do it manually ? when you 
# can iterate over a list; this is functional programming
# version2

# xnames will contain all the values we will pass to the function; 
# also if we repeat those values as the names on the cector,
# we will retain those names in the resulting list we are making
(xnames <- unique(iris$Species))
names(xnames) <- xnames

# map says to do the function for each item we give it and collect the results into a list
(my_list_of_result <- map(xnames,
                         \(x){do_one(x)
                         }))

# the results are there, to access them use them from the list i.e. 
my_list_of_result$versicolor

maybe I should turn this into a blog post, or canned response for this forum...

[saila]

Thank you so much for your response!!