Standard deviation for values below 1 When analyzing a dataset based on percentages, I have on occasions worked with data as "full" value (i.e. "50") or "reduced" value (i.e. ".50").
However, it just occurred to me that this could have a serious impact on how standard deviation and variance.
If my standard deviation and variance are above 1, the standard deviation will be smaller than the variance. But if they are below 1, the standard deviation will be bigger than the variance.
Does it mean it is good practice to never have a dataset resulting in a variance below 1? Or is there a way to account for this?
 A: 
If my standard deviation and variance are above 1, the standard deviation will be smaller than the variance. But if they are below 1, the standard deviation will be bigger than the variance.

Actually, this is not the case, because you're ignoring the units. Standard deviation of a percentage is measured in percent, while the variance is not.
It's like having a standard deviation of 20 cm (variance 400 cm$^2$) and then worrying about if you measure it in meters, that the variance (0.04 m$^2$) is smaller than the standard deviation (0.2 m). While 0.04 is less than 0.2, you're comparing apples and oranges - they're in different units!
So you can't say that the variance is bigger than or smaller than the standard deviation. They're not comparable at all.
Nothing is amiss: you can happily work with values above 1 or below 1; everything remains consistent. There's nothing to account for.
The only source of a problem I perceive there is if some of your values are in whole percent (50) and some are written as a fraction (0.50), without keeping track of the corresponding units. That you would worry about.
A: Variance (which ST is derived from) measures differences from the mean. Given differences from the average d1 and d2, whether D > 0 or D < 0 this always applies: d1 <=> d2 === d1^2 <=> d2^2. 
Examples:
d1 = 2
d2 = 3
2 < 3 and 4 < 9

d1 = 0.1
d2 = 0.5
0.1 < 0.5 and 0.01 < 0.25

d1 = 0.5
d2 = 2
0.5 < 2 and 0.25 < 4

I don't know exactly how standard deviation works in practice but it holds true that larger differences give a larger value and small differences give a smaller value even if squaring switches from increasing to decreasing going down passed one.
I think this is valid as once you reach as low as 1, the only way from there toward 0 is down. On a plotter it should produce a continuous curve.
Standard deviation will be compared to standard deviation and variance with variance. As long as you stick to that whether using the standard deviation (square root) or variance (averaged squared differences) they'll remain proportionately comparable.
I believe that SD will effectively be converting it from a geometric (squared curve) to a linear curve.


*

*Please excuse that I am from a programming background. The notation I use is works are <=> means compare (-1 less than, 0 same, 1 more than) and === means exactly equal. Numbers in variables are part of the variable name, multiplication only happens with an explicit *.

