How to interpret bctobit output

[Victoria Tanimonure]

I got this output after running bctobit command for specification test in tobit, how do I interpret it

LM test of Tobit specification
Bootstrap critical values
lm %10 %5 %1
5.2015 3.37393 4.930409 9.2035513

[William Lisowski]

Welcome to Statalist!

The lm statistic is 5.2015. The critical value at 10% is 3.37, at 5% is 4.93, and at 1% is 9.20. This is just like having, say, a t statistic and critical values for that statistic, except that for the lm statistic, the critical values are not calculated from tables of the distribution of the statistic under the null hypothesis, but rather by bootstrap techniques. Your lm statistic exceeds the 5% cutoff, which means that at the 5% confidence level you would reject the null hypothesis. Or put differently, the p-values for the test lies somewhere between .05 and .01.

For the LM test the null hypothesis is that the Tobit model is suitable, so this suggests that the Tobit model is not suitable for your data, at a p-value less than .05.

Now, I'll confess that I've not used bctobit, which by the way is a user-written command, not part of the official Stata release. The information presented above comes from reading the output of the help bctobit command. Further information on the test is available from the author of the command in http://www.stata.com/meeting/boston1...10_vincent.pdf . In particular I don't know if the alternative hypothesis used in bctobit is a sensible one, or if there are other more appropriate tests of the suitability of the tobit model to your data.