Dear all,
I'm a student and a new user of Stata. I have some problems with the following regression.
The purpose of my analysis is to see what is the relationship between age and the probability of being employed during the Great Recession. For this reason, I have created a binary variable called "young" (that is equal to 1 when the age of the worker is <25) and have used several control variables in a probit model.
Then, I've tried to analyze how could change the marginal effect of "young" when I consider the weekly earnings the
command I can see a difference in the probability of being employed between those who are young and those who are not, particularly high for the lower weekly earnings values.
However, when I create a new dummy variable "poor" (that is equal to 1 when "earnwke"<260) and insert the interaction term "youngpoor" (that is the product between "young" and "poor") in another regression, the interaction coefficient is not statistically significant (both if I use a probit model both if I use a linear regression model) and I can't understand why.
I attach the files that describe the variables used and a short do-file to be able to replicate my results. The data file contains data on 5412 workers who were survey in the April 2008 Current Population Survey and reported that they were employed. The data file contains their employment status in April 2009, one year later, along with some additional variables.
Thanks in advance for your reply
Here is the do file:
Code:
I'm a student and a new user of Stata. I have some problems with the following regression.
The purpose of my analysis is to see what is the relationship between age and the probability of being employed during the Great Recession. For this reason, I have created a binary variable called "young" (that is equal to 1 when the age of the worker is <25) and have used several control variables in a probit model.
Code:
*At first, I use only young variable generate young = age<25 probit employed young female married race ne_states so_states ce_states private educ_lths educ_hs educ_somecol educ_aa educ_ba union earnwke, robust
Code:
marginsplot
However, when I create a new dummy variable "poor" (that is equal to 1 when "earnwke"<260) and insert the interaction term "youngpoor" (that is the product between "young" and "poor") in another regression, the interaction coefficient is not statistically significant (both if I use a probit model both if I use a linear regression model) and I can't understand why.
I attach the files that describe the variables used and a short do-file to be able to replicate my results. The data file contains data on 5412 workers who were survey in the April 2008 Current Population Survey and reported that they were employed. The data file contains their employment status in April 2009, one year later, along with some additional variables.
Thanks in advance for your reply
Here is the do file:
Code:
Code:
*At first, I use only young variable generate young = age<25 probit employed young female married race ne_states so_states ce_states private educ_lths educ_hs educ_somecol educ_aa educ_ba union earnwke, robust margins, at((means) _all earnwke=(0 500 1000 1500 2000 2500 3000) young==(0 1)) marginsplot *Then, I use also poor and the interaction variable generate poor = earnwke<260 generate youngpoor = young*poor probit employed young female married race ne_states so_states ce_states private educ_lths educ_hs educ_somecol educ_aa educ_ba union poor youngpoor, robust *So, I've tried with a LPM regress employed young female married race ne_states so_states ce_states private educ_lths educ_hs educ_somecol educ_aa educ_ba union poor youngpoor, robust
Attached Files
