0
$\begingroup$

I'm trying to find a way to visualize the results of an association analysis where I corrected for confounding variables.

I have a set of cytokine data (amount of protein in the blood) from a set of patients infected and uninfected with HCV. The difference between the two is minimal when we test/visualize the raw data. However, when I adjusted for things like Race, gender, age, and disease status (by building a simple linear model) the p-values improved drastically.

However, I'm having trouble finding a way to show this visually. Normally I would just show a box-plot or bee-swam plot or something similar. But the raw data is unimpressive.

Does anyone have any ideas on ways to show this beyond reporting a p-value and the confounders/effect sizes?

$\endgroup$
  • $\begingroup$ Perhaps something along the lines of a partial residual plot (whether in the form of a boxplot or whatever)? $\endgroup$ – Glen_b -Reinstate Monica Feb 8 '13 at 22:58
0
$\begingroup$

How about a trellis plot? (If you are using R, see the lattice package, for one).

Another would be to do a parallel box plot, not of the actual data but of the predicted values from the linear model.

$\endgroup$
  • $\begingroup$ Thanks for the Trellis plot idea. It works well here because I can keep the same units across all of the plots. I find that biologists can wrap their heads around this easier. $\endgroup$ – JudoWill Feb 11 '13 at 15:35
1
$\begingroup$

If you have infected and uninfected individuals and a level of exposure (in this case Cytokine levels) you should be able to calculate a relative risk, and an appropriate 95% CI after adjustment. These can then be used in a plot that looks very similar to a boxplot if your levels are organized categorically, or a line diagram if your cytokine levels are more continuous.

For example, but with categorical levels rather than studies on the X-axis.

enter image description here

I'll post an example for continuous variables once I get off a train and have a working copy of R.

$\endgroup$
  • $\begingroup$ Interesting, and if I was showing this to a set of statisticians, mathmaticians, etc. this would certainly be a good way. But I was looking for a way that required less stats explanation. @Peter's Trellis plot is easier for a biologist to understand. $\endgroup$ – JudoWill Feb 11 '13 at 15:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.