Zero values of the dependent variable

[Shailaja Tiwari]

Dear all I have a question about using Zero values of the dependent variable.

I am using panel data at the intra household level. The data is collected for three rounds so time period is 3 and the panels are the individuals within the household. I want to see the impact of droughts, floods and price on individual food consumption.

For this I am running a regression (using mixed as my data is hierarchical) where the dependent variable is food consumption of individuals. As explanatory variables I include Literacy, Price and dummy variables for Droughts and Floods. Since my dependent variable is 'consumption of individual household members" by definition it can take the value of zero i.e., some member within the household did not
consume any food.
My question is that while running my regression should I use zero where some member did not consume anything or leave it as blank?

Thank you in advance

[Carlo Lazzaro]

Shailaja:
you should keep zero if zero is actually reported.
Missing values are Ok if you actually have unobserved values.

[Shailaja Tiwari]

Thank you for your reply Carlo Lazzaro.

[Shailaja Tiwari]

Dear Carlo Lazzaro if I use the log of my dependent variable (Dep var is food consumption in grams per person per day) should I then use zero?
For example if I have zeros in Consumption and i take the log of it to generate a new variable Log_Consumption then I will have empty cells in the places where Consumption was zero.
Should I replace these empty cells in Log_Consumption with zero?

[Carlo Lazzaro]

Shailaja:
no, it's missing, due to the log transformation.

[Shailaja Tiwari]

Dear Carlo Lazzaro if I understand it correctly

If I use the Consumption variable (without log transformation) then I retain the zeros while running my regression.
However, if I use Log_Consumption in that case i use the missing values (which will automatically occur due to the log transformation).

[Carlo Lazzaro]

Correct with an addition: if you go log, Stata will omit all the observations with missing values in the regressand (over and above omitting all the observations with at least a missing value in any variabe).

[Shailaja Tiwari]

Dear Carlo Lazzaro thank you for clarifying this issue.