1

Is there a way to exponentiate (ie, take antilog) of Stata's regression results table?

StatsScared
  • 1,218
  • 1
    "exponentiating the table" makes no sense at all. e.g. if you exponentiate a standard error, a t-value and a p-value, you'll get nonsense (the last won't be a probability for example). What are you actually trying to achieve? – Glen_b May 29 '19 at 05:57

1 Answers1

3

As @Glen_b said, you don't want to exponentiate all elements of the regression table.

However, you can exponentiate the coefficients and adjust the standard errors, p-values, confidence intervals accordingly. I assume from your question that you want to do this for linear regression. This can be done, and there are special cases where this makes sense. However, given the information you have given use we have no way of determining whether your problem falls in that special case. There is a discussion of how to do that, when this may be appropriate, and what the results mean in:

Roger Newson (2003) Stata tip 1: The eform() option of regress. The Stata Journal, 3(4):445. https://dx.doi.org/10.1177/1536867X0400300412

After reading that you can determine whether you really want to do this.

Maarten Buis
  • 21,005
  • +1. It's difficult to imagine applying this approach to anything but the intercept term, though, because it amounts to the relation $$\exp(\hat y)=\exp(\hat\beta_0+\hat\beta_1 x_1 + \cdots + \hat\beta_p x_p + \epsilon) = \exp(\hat\beta_0)\exp(x_1)^{\hat\beta_1}\cdots\exp(x_p)^{\hat\beta_p}\exp(\epsilon).$$ Notice that the only exponentiated parameter is the intercept $\hat\beta_0.$ – whuber May 29 '19 at 11:38
  • This was helpful. Thank you. With 'eform' I managed to do what I wanted: display the coefficients table in exponentiated form. However, it also transfoms SEs and CIs. This transformation is not a simple exponentiation of each corresponding SE and CI. Do you know how 'eform' transforms SEs and CIs? (The Stata Manual says nothing on this) – StatsScared May 29 '19 at 19:56
  • 1
    https://www.stata.com/support/faqs/statistics/delta-rule/ . this FAQ does not explicitly mention regress, but I suspect that this is what is used. You can easily check yourself. – Maarten Buis May 29 '19 at 20:05