Deutsches R-Forum

Code: Alles auswählen

# Modell laden
df <- read.csv2("Test_R.csv")

# Plottdarstellung und Vorhersage Modell 0
plot(df$Monate,df$y)
modell = nls(y~SSlogis(Monate,alpha,Wendepunkt,scale),data=df)
summary(modell)
predict(modell)
plot(df$Monate,df$y)
lines(df$Monate,predict(modell),col="blue",lty=2,lwd=3)
summary(modell)$coefficients[,1]

Code: Alles auswählen

for (i in 1:ncol(df)-1){
  modell= nls(i~SSlogis(Monate,alpha,Wendepunkt,scale),data=df)
  summary(modell)$coefficients[,1] }

Code: Alles auswählen

# Pkgs --------------------------------------------------------------------

library("tidyr")
library("dplyr")
library("purrr")
library("broom")
library("ggplot2")


# Load --------------------------------------------------------------------

df <- read.csv2("http://forum.r-statistik.de/download/file.php?id=1499")


# Clean -------------------------------------------------------------------

# bring to long format
dfl <- tidyr::pivot_longer(df, -Monate, names_to = "variable")


# Analyze ----------------------------------------------------------------------

# or (more flexible) in several steps
res <- dfl %>%
  group_by(variable) %>% # for each variable
  tidyr::nest() %>%
  mutate(
    # fit model
    fit = purrr::map(data, ~ nls(value ~ SSlogis(Monate, alpha,Wendepunkt,scale), data = .x)),
    # extract coefs from model   
    coefs = purrr::map(fit, broom::tidy),
    # compute gofs from model
    gofs = purrr::map(fit, broom::glance),
    # compute fitted values
    pred = purrr::map(fit, fitted)
  )

# we get for each variable a column
res

# all ceofs now in a nice data.frame
res %>%
  unnest(coefs)

# or the gofs for all models in a dat.frame
res %>%
  unnest(gofs)

# access individual models
res$fit[[2]]

# plot  all models
res %>%
  unnest(c(data, pred)) %>%
  ggplot(aes(x = Monate)) +
    geom_point(aes(y = value)) +
    geom_line(aes(y = pred)) +
    facet_wrap(~variable)