CODEWITHSUNDEEP
r
Maximilian
Maximilian

Reputation: 4229

How to re-scale logit output values

This is going into statistics, however quite few question has been answered here on SO. If there will be objections I will remove this question asap.

I would like to re-scale logit values so that the output values are scaled so that the min value is 0 or close to 0 and max value close to 1.

Here is sample data

logit <- rlogis(100, location = 1, scale = 5)

y <- sort((1/(1+exp(-logit))))[40:(length(logit)-50)] # here removing the close to 0 and 1 to replicate the problem.
x <- 1:length(y)

data <- data.frame(y,x)

with(data, plot(y~x))

So would someone point me to r package or show otherwise how to approach this and re-scale these values? Thanks.

Upvotes: 2

Views: 327

Answers (1)

Ben Bolker
Ben Bolker

Reputation: 226662

You just want the min and max values to have large absolute values.

out_range <- c(-10,10)
L_sc1 <- (logit-min(logit))/diff(range(logit)) # rescale to [0,1]
L_sc2 <- out_range[1]+diff(out_range)*L_sc1    # rescale to (min,max)

Plot results:

plot(sort(plogis(L_sc2)))

Upvotes: 2

Related Questions