Sometimes a continuous outcome variable does not follow a normal distribution. It could be skewed to the left or right. Either way our analysis is limited by the fact that the variable does not have a bell-curved symmetry distribution. I guess we can throw out all our work. After spending weeks, months, years of gathering this data which is now garbage.So sad. .. Or maybe we can put the variables in a format that follows a normal distribution.
A log transformation is simply taking the log of your continuous variable so you are dealing with data on a lower scale. If your values are in the tens of thousands like viral load data, the log transformation comes in handy. Here is an example in SAS programming:
data one; *create a new file named one;
set virol; * get the viral load data;
logviral=log(vload); *here we obtain the log transformation;
run; *execute your program;
The variable , logviral, is the log transformation of “vload”. Once you generate your new variable it can be inserted in your parametric models for further analysis.
If you enjoyed this short blog there are more the follow in the days to come…-Amy