I am trying to fit a log-linear model to a large number of variables from survey data. There are some reasons that it might be preferable to fit logistic regressions to that data instead. Several authorities suggest that these are equivalent. However, I have some reasons to doubt this.

  1. Log-linear models treat all variables equivalently, while logistic regression requires that one variable be identified as the response variable.
  2. In the least-squares context, it is not generally the case that for Y = a+bX + ε versus X = c+dY + ε the parameter d is even approximately equal to 1/b. This is because the first equation minimizes vertical error, while the second minimizes horizontal error. These will be equal only if the errors are symmetrical around the estimated line. So I am worried that this might be true of the logistic regression also. (2) is actually just a specific form of (1), i. e. a possible asymmetry in the regression format from the choice of a particular variable as response.

  3. If all the variables in the log-linear model are involved in one or more interaction terms, I do not see how a logistic regression can be equivalent. How would one express the interactions in which the response variable is involved in the context of a logistic regression?

In response to Bill Huber, I am using the term log-linear model in a considerably more restricted sense than Wikipedia. I am referring to models of categorical or ordinal count data, arranged in tables, where the coefficients are the total table count, the marginal counts for each factor divided by the total table count (serving as proxies for probabilities), and various interaction terms. This is the sense used in Agresti, “Categorical Data Analysis”, among others.

  • $\begingroup$ My understanding of "loglinear model"--which is pretty well aligned with the Wikipedia definition, albeit a bit more general--does not enable me to make sense of this question. Could you please tell us what this term means to you? $\endgroup$
    – whuber
    Commented Mar 27, 2013 at 22:54
  • 2
    $\begingroup$ Wikipedia calls it loglinear analysis: en.wikipedia.org/wiki/Loglinear_analysis $\endgroup$ Commented Mar 27, 2013 at 23:46
  • 1
    $\begingroup$ There several forms of loglinear analysis - general, logit, etc. Logit loglinear analysis is for situation when you have a dependent variables and some predictors. As far as I know, it gives the same result (parameter estimates) as nominal logistic regression. $\endgroup$
    – ttnphns
    Commented Mar 28, 2013 at 7:54

1 Answer 1


The answer is 'no'. The loglinear model is more general than the logistic regression model. See Fienberg, 1980, Analysis of Cross-Classified Categorical Data, section 6.2 on how to specify a loglinear model so that it corresponds to logistic regression.

Actually the reverse is true: If all variables are categorical, then every logistic regression model corresponds to some loglinear model.

  • $\begingroup$ So, this is what I thought, but not what I was hoping for. In R, which is the main tool I use, there are several contributed packages or available routines to adjust for complex survey sampling design for the lm and GLM models. I have not found anything equivalent for log-linear models. $\endgroup$
    – andrewH
    Commented Apr 2, 2013 at 23:28
  • $\begingroup$ I don't know whether it's useful for your purposes, but the loglm() function from MASS can fit log linear models. Also, you might like to check out the ACD package. $\endgroup$ Commented Apr 3, 2013 at 9:33
  • $\begingroup$ Also, with a Poisson GLM you can fit log-linear models. See Agresti: Categorical Data Analysis Section 8.6.7. $\endgroup$
    – Momo
    Commented Apr 3, 2013 at 10:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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