Quickly accessing p-values and confidence interval limits

[Dan Bratton]

After running a regression model in Stata a quick way to access the coefficients and their standard errors is using _b[varname] and _se[varname]. However, to access their corresponding p-values and confidence interval limits I find myself having to create a matrix of the results from which to extract them, like so:

sysuse auto, clear
regress length weight
mat R = r(table)
local p = R[4,1]
local ll = R[5,1] etc...

The code is rather ugly, having to manually specify the entries of the matrix from which to extract the relevant results. Also, when using multivariable regression I often end up with a sea of numbers from which I need to find my statistics of interest. Clearly this would be much easier (particularly when writing an ado file) if I could instead simply type

local p = _p[weight]
local ll = _ll[weight] etc...

in a similar manner to the _b[] and _se[] functions. My question is therefore, is there an easier way to extract p-values and confidence interval limits to what I have been using, and if not, might Stata consider adding functions like _p, _ll, _ul, _df etc in a future Stata release?

Thanks,

Dan

[Attaullah Shah]

Try something like

local t = _b[_cons]/_se[_cons]
local p =2*ttail(e(df_r),abs(`t'))

[Dan Bratton]

Thanks. I'm aware that that is another way of doing it but it still wouldn't be as easy as simply specifying _p[_cons] or _ll[_cons] etc. I realise that this is only a minor gripe of mine but I do not understand why you can use _b and _se to obtain the coefficient and its standard error but not similar things for p-values, CIs etc.

[Attaullah Shah]

You can use the above code to write your own program with rclass properties.

[Steve Samuels]

I used to use feed r(table) into Mata to manipulate estimation results, until I discovered Roger Newson's parmest package (SSC).