I have a dependent variable that is essentially percent changes from a previous period. Since the data could range between negative and positive values, I am looking at ways that would best treat or transform such data. Thanks in advance!
Data looks something like this
Measure / percentage from last period
4 / +25%
3 / -10%
1.5 / +35%
2 / -22%
5 / +7%
6 / +3%
6.5 / -0.5%
7 / -22%
Hello Sumiran,
Taking a series of measurements as % from one period to the next creates a new variable with a variance structure that differs from the measured variable. Which makes it difficult to interpret the result of the analysis. And, DVs as a % can ugly, with respect to the usual assumptions of normal error that underpin many statistical analyses.
For what it is worth consider the follow algebra.
lnY = a + b* time
exp(lnY) = exp(a) * exp(b * time)
Y = exp(a) * exp(b * time)
You might recognized this as an exponential equation.
A common application is the exponential growth rate b,
which has units of %/time.
So an estimate of b might be appropriate, depending on the purpose of the analysis.
~David S
Hi. You haven't said what you want to do with the data. The answer depends on that. A graph or two is also usually helpful.
Best wishes, David
Hello Sumiran,
Taking a series of measurements as % from one period to the next creates a new variable with a variance structure that differs from the measured variable. Which makes it difficult to interpret the result of the analysis. And, DVs as a % can ugly, with respect to the usual assumptions of normal error that underpin many statistical analyses.
For what it is worth consider the follow algebra.
lnY = a + b* time
exp(lnY) = exp(a) * exp(b * time)
Y = exp(a) * exp(b * time)
You might recognized this as an exponential equation.
A common application is the exponential growth rate b,
which has units of %/time.
So an estimate of b might be appropriate, depending on the purpose of the analysis.
~David S
- Badges
- Science topic