Median Absolute Deviation in R isnt like in class

[RYann]

Hey y'all
In class we learned that

48, 52, 76, 88, 92, 92, 100
AD=15.43
but then in R i tried and i get
new <- c(48, 52, 76, 88, 92, 92, 100)
mad(new)
17.7912
Any thoughts?

[focardozom]

For some reason I don't understand, they use a constant value (1.4826) to ensure consistency. I think you can change that parameter (constant=?) to get the class MAD.

mad function | R Documentation

[EconProf]

MAD is the median absolute deviation from the median. Your formula from class is for the mean absolute deviation from the median. The simple median value is 12, which when multiplied by the default constant of 1.4826 for asymptotically normal consistency gives the reported value of 17.7912.

Some general advice: when something appears to be wrong, check the R help documentation.

[RYann]

Thank you sir! Is there a command or an option to change R settings (in a specific syntax if needed) so it calculates the mean instead of the median?

[EconProf]

mean(abs(new-median(new)))

mean(), abs() and median() are all functions in the stats package

[RYann]

@EconProf

mean(abs(new-median(new)))

thanks. also helped me with finding
mean((new-mean(new))^2) #mean diff

[Yarnabrina]

Hi @RYann, a lot of statistical functions are already there in R. To calculate variance, you can use var. You don't need to calculate manually.

However, I should add, most of these will calculate sample versions (which will be unbiased estimators of population versions) only, so for var, n-1 will be used as the denominator instead of n. If you specifically need population version, then your solution is correct way.

Hope this helps.

[RYann]

Cheers man, I appreciate any tips! I'm only starting in this world. So any help is much appreciated.

I use var like you say,
Var(sample(X,10000, replace = T, prob=Px)
Our prof. showed us the basic principle that the bigger the sample, the closer var gets to its "pure var". Also using table(X)/length(X)...

Happy new year