Tobit (if the values of dependent variable ranges between -6.96296 + 18.89634

[Olive Bat]

Tobit (if the values of dependent variable ranges between -6.96296 + 18.89634 what should be command in STATA for Tobit be..

tobit sav_index for_remitt_hh DRatio i.Locality_hh i.sex_hh i.age_group_hh i.edu_hh i.eco_activity_hh i.married_hh i.hhsize i.religion_hh i.District_Code i.land_acresnew , ll(-1)..is the above command right.

I would like to know more details on Heckman selection model which could be opted instead of the above tobit.
what are its command on STATA and what do we opt for it instead of tobit. I ran a tobit as my expenditure values concentrated between 0 and 1 .. so created an index so a negative value came into picture.. (-6.96296 + 18.89634.. this is the range )

please clarify .

[Phil Bromiley]

Heckman is only used if you have selected/missing data. It looks like you're artificially bounding the data (creating problems that didn't exist in the original data). Tobit only makes a difference if there is a boundary (e.g., values can't be less than zero so anything that would have been below zero is coded as zero). But, it doesn't sound like your data has such a boundary. Even if it were bounded to be between -7 and 19, without any values at -7 or 19, I don't see the point of tobit.

I have no idea what you're saying about the new range. Why create an index just to rescale stuff? If expenditures really are bounded by 0 and 1 (and you actually get 0 and/or 1 in the data), then tobit does sound right. But working with the original data is easier than working with rescaled data - predicted values in the original data have meaning whereas they are hard to interpret in the rescaled data.

[Jeff Wooldridge]

Olive, you need to provide more information. Like Phil, I’m puzzled by what you did to your y variable. Did you start with a variable in the interval [0,1]? If so, do you see zeros or ones? We can help if this is your setup.