"bys industry fyear: asreg var1 var2 var3, fit" create variables having no values (missing)

[Zihao Zhou]

the industry is a two digit number of sic (numerical), fyear is the financial year.
I use "bys industry fyear: asreg var1 var2 var3, fit", and the resulted "_Nobs; _R2; _adjR2; _b_var2; _b_var3; _b_cons _fitted; _residuals" all have no values (missing values in all variables). But when I do not add "bys industry fyear" and only use the asreg function, then the created variables have values. I do not know how to solve this problem
Thanks

[Nick Cox]

How many observations do you have for each industry, year pair? Perhaps you have this problem:

. webuse grunfeld

. bysort company year : regress mvalue invest kstock

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1935
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1936
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1937
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1938
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1939
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1940
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1941
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1942
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1943
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1944
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1945
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1946
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1947
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1948
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1949
insufficient observations

---------------------------------------------------------------------------------------------------------------
-> company = 1, year = 1950
insufficient observations

Trying to fit a plane in 4 dimensions won't work if you have (as here) just one data point in each regression and in fact you need a lot more than one for a good result.

[Zihao Zhou]

Thanks, the code is conducted on a sample data with only 200 observations, I will increase the sample size significantly and see what happened.

[Jared Greathouse]

Why is this necessary?

 bys industry fyear