Subpopulation treatment effect pattern plot (STEPP)
stepp_tail regression_cmd [yvar] zvar [adjvars] [if] [in] , options
stepp_window regression_cmd [yvar] zvar [adjvars] [if] [in] , options
stepp_plot stubname [, options]
regression_cmd may be clogit, cnreg, glm, intreg, logistic, logit, mlogit, nbreg, ologit, oprobit, poisson, probit, qreg, regress, stcox, streg, or xtgee.
options Description ------------------------------------------------------------------------- gen(stubname) creates five new variables containing results of STEPP analysis g(#) (stepp_tail only) defines the number of subpopulation groups n1(#) (stepp_window only) defines the number of individuals belonging only to one of two neighbouring subpopulations n2(#) (stepp_window only) defines the number of individuals in a subpopulation treatment(trt_varlist) defines the list of variables whose interactions with zvar are to be studied regression_cmd_options options for regression_cmd
options for stepp_plot vn(#) variable number in treatment() plot(plot) adds other plots to the generated graph graph_options options of graph twoway -------------------------------------------------------------------------
All weight types supported by regression_cmd are allowed; see weight.
yvar is not allowed for streg and stcox. For these commands, you must first stset your data.
ststep_tail and ststep_window compute Bonetti & Gelber (2000, 2004)'s STEPP estimators for graphical exploration of a treatment/covariate interaction. ststep_tail provides the tail-oriented estimator, and ststep_window the sliding-window estimator. Plotting the results may be done by using stepp_plot, in which case stubname is the same as in the gen(stubname) option of stepp_tail and stepp_window.
zvar is the continuous covariate whose interaction with treatment is to be studied, and adjvars is a list of other covariates used to linearly adjust each model fitted to the treatment variable(s) defined by treatment().
Options for stepp_tail and stepp_window:
gen(stubname) creates five new variables called stubnameb, stubnamese, stubnamemean, stubnamelb, stubnameub. stubnameb is the estimated regression coefficient in each subpopulation, stubnamese is its standard error, stubnamemean contains the mean of zvar in each subpopulation, and stubnamelb and stubnameub are pointwise 95% confidence limits for stubnameb. If treatment() includes more than one variable, the created variables have 2, 3, ... appended to the names, e.g. stubnameb2.
g(#) (stepp_tail only) defines the number of subpopulation groups. The actual number of subpopulations used is 2 * # - 1.
n1(#) (stepp_window only) defines the number of individuals belonging only to one of two neighbouring subpopulations.
n2(#) (stepp_window only) defines the number of individuals in a subpopulation. The overlap between two neighbouring subpopulations is n2() minus n1() individuals.
treatment(trt_varlist) defines the list of variables whose interactions with zvar are to be studied. Typically trt_varlist will comprise just one binary variable, representing the two arms of a parallel-group clinical trial.
regression_cmd_options are options for regression_cmd.
Options for stepp_plot:
vn(#) # is an integer defining the variable number in treatment(), when more than one variable is specified. When only one variable is specified, vn() is not required.
plot(plot) provides a way to add other plots to the generated graph; see help plot option.
graph_options are options of graph twoway, such as xtitle(), ytitle(), etc.
. stepp_tail regress y x a1 a2, g(10) gen(z) treatment(t)
. stepp_window stcox x a1 a2, n1(40) n2(50) gen(z) treatment(t)
. stepp_plot z, xtitle("Serum rhubarb") ytitle("log relative hazard") name(myplot)
Patrick Royston, MRC Clinical Trials Unit, London. email@example.com
M. Bonetti and R. D. Gelber. 2000. A graphical method to assess treatment-covariate interactions using the Cox model on subsets of the data. Statistics in Medicine 19: 2595-2609.
M. Bonetti and R. D. Gelber. 2004. Patterns of treatment effects in subsets of patients in clinical trials. Biostatistics 5: 465-481.
P. Royston and W. Sauerbre. 2009. Two techniques for investigating interactions between treatment and continuous covariates in clinical trials. Stata Journal 9(2): 230-251.
W. Sauerbrei, P. Royston and K. Zapien. 2007. Detecting an interaction between treatment and a continuous covariate: a comparison of two approaches. Computational Statistics and Data Analysis 51: 4054-4063.
Article: Stata Journal, volume 9, number 2: st0164