How to deal with true zeros in our database?

[Afonso Rodrigues]

Hello

I'm doing a research where I (amongst others) use a variable that represents the nuclear energy consumption. The problem is that I have some countries who do not have nuclear energy and others that only started using this kind of energy in the middle of my time period. So, all this values are zeros. My question is, how can I treat this zeros on Stata? Is there something special I need to do? Is it possible to even use these countries?

I appreciate all the help

[Nick Cox]

Zeros are zeros; no problem there. Stata can handle zeros.

What you're really pointing towards, I imagine, is some question about statistical analysis or modelling and you're going to have to tell us more about what you intend. . For example, some questions refer to countries with nuclear energy and those with none are irrelevant. Some questions refer to all countries.

[Carlo Lazzaro]

Afonso:
as an aside to Nick's wise reply, the issue seems to refer to the data generating process:
- soime countries where the nuclear power is available, can decide not to use it (and this is one sort of zeros);
- some countries simply do not have nuclear power available (and this is another sort of zeros).
As Nick pointed out, you should provide the list with more details about what you're experiencing with your dataset.

[Afonso Rodrigues]

Thanks for you answers. I have more energy variables than only the nuclear one, so that's why I have in my database countries that do not use this kind of energy. I am in the first stages of my study, so I still can't give you many insights about the model and every econometric aspects of my research. But, a priori, one of the problems I anticipate with this true zeros thing is, what if I need to logarithmize the variables? (not sure if that's the right way to spell it) It won't be possible to do it where the values are zero.

[Carlo Lazzaro]

Afonso:
logging (handy as it is, especially in log-linear regression model) comes at the cost of having missing values when data are zeros in their original metric.
It's up to the researcher (but the literature in your reserch field can probably give you some suggestion) to decide between bearing the missing values (and so reducing the -e(sample)- included in the regression model) or keeping the variables in their original metric.