# Thread: Trichotomizing an Independent Variable

1. ## Trichotomizing an Independent Variable

In SPSS, how does one go about trichotomizing an independent variable such that about 33% are in each of the three groups?

2. ## Re: Trichotomizing an Independent Variable

Are you looking to just create tertiles or are you looking to break a continuous variable up based a prediction, e.g., if your value of X is < ?? then your risk of Y is ??; value between ?? and ?? then risk of Y is ??; and if value > ?? then risk of Y is ??

3. ## Re: Trichotomizing an Independent Variable

Originally Posted by hlsmith
Are you looking to just create tertiles or are you looking to break a continuous variable up based a prediction, e.g., if your value of X is < ?? then your risk of Y is ??; value between ?? and ?? then risk of Y is ??; and if value > ?? then risk of Y is ??
Good questions-- very important to have a theoretical and empirical reason to categorize a continuous variable.

4. ## Re: Trichotomizing an Independent Variable

Originally Posted by langood
In SPSS, how does one go about trichotomizing an independent variable such that about 33% are in each of the three groups?
Don't - it just reduces power. If the IV is continuous, you don't need to split it up to run an appropriate analysis. See Cohen 1983: conservancy.umn.edu/bitstream/handle/11299/107497/v07n3p249.pdf?sequence=1

5. ## Re: Trichotomizing an Independent Variable

I've also read something that said if you are hoping to generalize ***tiles, it may be difficult to generalize effects since textiles in a new sample may not be the same numbers due to heterogeneity between samples. Though they can know your distribution better if you provide moments.

I guess another reason to trichotomize data is to get three random subsample for say a training, test, and validation set.

 Tweet