logarithmic linear fit plotting in r

Question

I have a dataframe correspond on frequency of each variable (example:variable 1 appear 1984 times and variable 2 appear 974 ...)

dff<-data.frame(Var1=c(1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 ,9 ,10 ,11, 12 ,13 ,14 ,15 ,16 ,27, 30 ,35 ,36 ,38 ,39 ,40 ,41), Freq=c(1984,  974,  464 , 251 , 127 ,  83 ,  45 ,  26  , 16  , 12 ,   9   , 5 ,   5  ,  2    ,3  ,  1   , 1  ,  1 ,   1   , 2  ,  5,4,2,1))

plot(dff$Var1,log(dff$Freq))

log plot

as we see in the picture, there is a linear regression,

I just want to find a method that can fit only the regression part where the linear fit intersect the x axis.

because, I need to extract the other points that are not fitted in the linear fit (points after dff$Var1=20) that is mean is not a noise point

Maybe I am confused by the question, and the problem, but why don't you just drop the `>20` values and estimation a model with `lm`? Is that what you are asking? — Jason Morgan, Feb 09 '17 at 17:09
This statement is unclear "as we see in the picture, there is a linear regression,". Did you mean a linear relationship, or a correlation or something? — Hack-R, Feb 09 '17 at 17:12
because this is a small part of data, the dff dataframe is just a dataframe from a big dataframe of dataframe, that's mean I have like 1000 datatracks each datatrack has a dataframe, for that the number will differ from dataframe to another, and just I have to create a script that loop all dataframe and axtract the number — user3744999, Feb 09 '17 at 17:14
@user3744999 datatracks? That's probably a new term for most people. What does it mean? On Google it only shows up as a brand name, i.e. http://www.datatracks.com — Hack-R, Feb 09 '17 at 17:19
okay, I am sorry, but my question is not related to datatracks(which is just a series of data) , I just need a way to determine where the linear fit intersect the x axis — user3744999, Feb 09 '17 at 17:31

score 0 · Accepted Answer · answered Feb 09 '17 at 19:12

What about this:

plot(dff$Var1, log(dff$Freq))
lr <- lm(log(Freq) ~ Var1, data = dff[dff$Var1 < 20, ])
abline(lr)

The cutoff point is 20. But you can vary it according to what you are doing.

If you want to calculate

where the linear fit intersect the x axis.

Get the coefficients:

coef(lr)
(Intercept)        Var1 
  7.4636699  -0.4741615

And solve the equation 7.4636699 + Var1*(-0.4741615) = 0.

logarithmic linear fit plotting in r

1 Answers1