3 ms·
> Apparently (based on the comment I was replying to), labcomputer does seem to think the maximum is an outlier Visually, 1968 sure looks like an outlier to me
by labcomputer 5y ago
> Apparently (based on the comment I was replying to), labcomputer does seem to think the maximum is an outlier
Visually, 1968 sure looks like an outlier to me: https://fred.stlouisfed.org/graph/?g=EZn4 https://fred.stlouisfed.org/graph/?g=EZn4
But let's check my intuition. The federal minimum wage was raised 10 times from 1960 through 1980. The raises have the following indexed values:
* 1961: 137
* 1963: 145
* 1967: 151
* 1968: 167
* 1974: 146
* 1975: 143
* 1976: 147
* 1978: 150
* 1979: 151
* 1980: 141
So mean = 148, st.dev = 8.1. That puts 1968 at 2.4 standard deviations above the mean.
If you just want to make the pedantic point that the maximum is not in general an outlier, we are in agreement. But, in this particular case, the maximum is clearly an outlier.
- mannykannot 5y agoI appreciate the work you put into this, but when I look at the data, I see a nonstationary (humped) time series, where I am not sure that 2.4 standard deviations above the mean is a reliable identifier of outliers. Furthermore, I think the data still supports janandonly's point (though not quite as starkly) even if we throw out the maximum. It would indeed be pedantic of me to now make an issue of the maximum not in general being an outlier, but at the time I wrote that, I was replying to a post in which the only justification given for considering 1968 to be an outlier was the fact that it was the maximum.