I am populating a list of predicted values for every individual using regression equations.
where , Ln( predicted value ( girls))= -9.989+(2.018*Ln(ht))+(0.324*Ln(age))
I tried to populate the predicted values using the statement
"gen predvalue=0";
" predvalue= 2.718^( -9.989+(2.018*ht)+(0.324*age)) if sex==2"
However, the value of predvalue in this case, which should be closer to observed values, is very very low. For instance where observed value is close to 2.25, the predicted value is close to 0.25. The expected predicted values close would ideally be slightly higher ( 2.8- 3 etc). The low predicted values are leading to a distortion in percentage of predicted attained.
I cannot figure out the problem with my syntax
Yes, Tunc, your'e righy.
If we take the equation
Ln( predicted value ( girls))= -9.989+(2.018*Ln(ht))+(0.324*Ln(age))
and we are going to get rod of Ln from the left side then we have
e^Ln( predicted value ( girls))=predicted value( girls)
The same operation for the right side is
e^[-9.989+(2.018*Ln(ht))+(0.324*Ln(age))]=e^[-9.989]*e^[(2.018*Ln(ht))]*e^[(0.324*Ln(age))]=
e^[-9.989]*[e^Ln(ht)]^2.018*[e^Ln(age)]^0.324=e^[-9.989]*ht^2.1018*age^0.324
The above is the power model. And the equation
predvalue= 2.718^( -9.989+(2.018*ht)+(0.324*age))
is the exponential model which is equivalent to
2.718^( -9.989)*[e^(ht)]^2.018*[e^age]^0.3 - the expontial
model.
A two step conversion worked my case.
For ln p= ax+by+z, generated a new variable f=ax+by+z and got my predicted value using predvalue=e^f
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology