1. ## Different signs

Hello everybody:

I have a litte problem interpreting my regression output.

In the first regression, I have a continuous variable with a positive sign (there are further included some controls). Now, I created a median-dummy of this variable, means 1 if above-median, 0 otherwise. However in this regression with the same controls , the sign is negativ.

Now I don't understand why. If an increase of the continuous variable has a positive impact, the above-median observations should have the same sign, or am I wrong?

Thank you!

2. ## Re: Different signs

Hello, first of all just a comment regarding the "median-split" approach: Using the median to create a categorical variable from a continuous one is poor statistical practise and connected to several problems, as summarized e.g. in the paper of Macmallu, Zhang and Rucker (2002). Regarding the problem with the sign: Are all other variables the same in the two regression model? Could you provide some of the data or outputs? In priciple you are right, the sign should be the same, as long as the "0" is the baseline category in the factorial regression model...

3. ## Re: Different signs

Does the model know it is a categorical variable? It is always best to plot your data to understand things visually as well - before and after the split.

4. ## Re: Different signs

Thank you both for your reply. I just found my mistake, I had an additional variable in my dummy-regression that leads to the change in sign. But the Zhang and Rucker (2002) paper is very interesting!

Unfortunately, I have a new but similar question

This time, I have a variable that is calculated as the natural logarithm of debts/total assets (=leverage). The sign in the regression is always positive that supports my hypothesis. However, without the ln, the sign changes to negative and I again cannot evaluate if this is usual or why this happens?

5. ## Re: Different signs

2 questions :

Why did you transform the variable?

If you look at the distribution of the transformed variable (what do the values look like (positive, negative, a mix) ?

6. ## Re: Different signs

1. I first followed a paper that uses the natural logarithm. However, yesterday I recognized that it's definitely more common not to use the ln in literature, but this completely changes my results.

2. It's a mix: the mean is -0.73, median -0.62, p25 -1.05, p75 -0.33, min -7.11, max 5.09, sdt dev 0.69

Does that help?

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts