standardized coefficients from glm logit I am trying to create a coefficient plot from multiple logistic regression models, which all have the same predictors, but different sample sizes. This is a pre-test to a multilevel model. My question is two fold:


*

*Given, I want to compare the effect sizes of the same predictor in the different models, I assume I need to use standardized coefficients. How does one calculate standardized coefficients in a logit model?

*Is there an easy way to estimate such coefficients in R? For instance with OLS, I could rely on the "lm.beta" function from the QuantPsyc package. I am wondering, is there a functional equivalent for a glm logit? I could not find an immediate solution myself.
 A: If the predictors you used in the different regression models are measured in the same way it is no longer necessary to normalize the coefficients, since the predictors are already on the same scale the coefficients are as well.
If not, first try to rescale the predictor variables.
A: It depends on the scale that was used to measure the variables.
For instance, if your dependent variable is a logit that assumes a value of 0 and 1, and let's say you have a very large scale for an independent variable, e.g. 1,000,000, then the two scales are different and any interpretation of the output might not be accurate since the two scales are not in synchronization.
Therefore, I would be inclined to go back and see how your variables were scaled in the first instance. If they are significantly different, then scaling and then observing the resulting output could be warranted.
As regards your question on how to do this in R:
1) You could try to scale your dataframe by using the scale(df) command.
2) If you wanted to do this manually, you can also use what is called max-min normalisation. For a variable X, this would be calculated as follows:
X_Scaled = (X-min(X))/(max(X)-min(X))
Hope the above helps. As always, the theoretical basis behind your model is the best indication as to whether scaling of variables is warranted or not.
