# How skewed does a distribution have to be before we use the median as the best measure of central tendency? [duplicate]

The question is in the title. I'm mostly asking because I'm wondering whether this is subjective or not (this is regarding a sample like avg engineer salaries by state).

I know it's more appropriate to use the mean if the distribution is symmetric, or normal. But wouldn't all real life distributions be skewed, technically? What's a good rule of thumb here? When should I say, skewness value of $\pm \, x$ so the median is more representative of the average salary?

• stats.stackexchange.com/questions/9987/… – David May 25 '17 at 21:07
• If you're after "average" salary, you're better off looking at the median - that would be a question of what could the average person expect to earn, rather than what is the average of the salaries. – MotiNK May 25 '17 at 22:10
• The question is only resolved by clearly figuring out what you even mean by the vague phrase "central tendency" in your particular case. This is generally resolved by considering in explicit terms what you want to find out from your data. [If on the other hand your aim is simply to offer some quick summary of the data don't restrict yourself to any single measure; if means and medians and modes are substantively different, that's an important thing to know in a summary.] – Glen_b May 25 '17 at 23:13