{smcl} {* *! version 1.0.1 2023-03-23}{...} {vieweralsosee "stpm3" "help stpm3"}{...} {vieweralsosee "stpm3 extended varlist" "help stpm3_extfunctions"}{...} {vieweralsosee "stpm3 predictions guide" "help stpm3_predictions"}{...} {vieweralsosee "standsurv" "help standsurv"}{...} {viewerjumpto "Syntax" "stpm3km##syntax"}{...} {viewerjumpto "Description" "stpm3km##description"}{...} {viewerjumpto "Options" "stpm3km##options"}{...} {viewerjumpto "Remarks" "stpm3km##remarks"}{...} {viewerjumpto "Examples" "stpm3km##examples"}{...} {marker syntax}{...} {title:Syntax for stpm3aj} {pstd} Compare model based marginal predictions to non-parametric estimates after fitting a {helpb stpm3:stpm3} model. {pstd} By default only data included in the model is used. If you want to include data not in the model then use then {cmd:noesample} option. This can be useful for external validation. {p 8 16 2} {cmd:stpm3km} [{it:varname}] {ifin} [,{it: options,}] {p 8 16 2} {cmd:stpm3aj} [{it:varname}] {ifin} [,{it: options,}] {synoptset 25 tabbed}{...} {syntab:{bf: Options}} {synoptline} {synopthdr:option} {synoptline} {synopt :{opt cut(options)}}the breaks for the groups {p_end} {synopt :{opt noesamp:le}}predict out of sample{p_end} {synopt :{opt fac:tor}}states {it:varname} is a factor{p_end} {synopt :{opt failure}}calculate failure rather than survival{p_end} {synopt :{opt fr:ame(framename)}}predictions will be saved to frame {it:framename}{p_end} {synopt :{opt nogr:aph}}do not plot graph{p_end} {synopt :{opt gr:oups(#)}}Number of groups for continuous covariates{p_end} {synopt :{opt nokm}}Do not include Kaplan-Meier estimates on graph{p_end} {synopt :{opt legoptions(options)}}additional legend options{p_end} {synopt :{opt maxt(#)}}maximum follow-up time{p_end} {synopt :{opt ntime:var(#)}}Number of timepoints for standsurv{p_end} {synopt :{opt pit:ime(#)}}Time for prognostic index when time-dependent effects{p_end} {synoptline} {pstd} Additional options for {cmd:stpm3aj} {synoptset 25 tabbed}{...} {syntab:{bf: Options}} {synoptline} {synopthdr:option} {synoptline} {synopt :{opt crmodels(models)}}competing risks models{p_end} {synopt :{opt compet1(#)}}The failure value of teh first comoeting risk{p_end} {synopt :{opt compet2(#)}}The failure value of the second comoeting risk{p_end} {synoptline} {marker description}{...} {title:Description} {pstd} {cmd:stpm3km} can be used after fitting a {cmd:stpm3} model. Continous covariates are categorized into groups and the marginal survival or failure function is calculated within each group level and compared to the corresponding Kaplan-Meier estimate. {pstd} {cmd:stpm3aj} is similar but for competing risks models where model based estimates of the cumulative incidence function (CIF) are compared to non-parametric Aalen-Johansen estimates. The cause-specific models are passed to {cmd:stpm3aj} using the {bf:crmodels()} option with values of competing events passed using {bf:compet1(}{it:#}{bf:)}..{bf:compet}{it:k}{bf:(}{it:#}{bf:)} in the same way as {help stcompet}. {pstd} Note that the user written {help stcompet} command is used to obtain the non-parametric estimates of the CIF and so needs to be installed from SSC. {phang} If {it:varname} is not specified then the prognostic index is categorized. Use factor variable notation, e.g. {cmd:i.sex} to denote that a variable is categorical. Alternatively you can use the {cmd:factor} option. {phang} The default number of groups is 5. This can be changed with the {cmd:groups(#)} option. The groups will be of roughly equal frequencies. You can define your own groups using the {cmd:cut()} option. {marker options}{...} {title:Options} {phang} {opt cut(numlist)} a numlist in ascending order giving the breaks for the groups of {it:varname} or the prognostic index. {phang} {opt noesample} allows data not included in the model to be used. This is useful for external validation. {phang} {opt factor} specifies that {it:varname} is a factor (categorical) variable. {phang} {opt frame(framename)} saves the Kaplan-Meier and model based standardized estimates to a frame. {phang} {opt nograph} suppresses plotting of the graph. It only makes sense to use this option in conjunction with the {cmd:frame()} option. {phang} {opt groups(#)} specifies how many groups to create. The default is 5. If there are many ties in {it:varname} there may be less groups then specified. {phang} {opt nokm} exclude Kaplan-Meier estimate from, graph. {phang} {opt legoptions(options)} additional legend options. This may help with positioning size etc. For full control, you can supply your own {cmd:legend()} option. If doing this {cmd:legoptions()} is ignored. case {phang} {opt maxt(#)} specifies the maxiumum follow-up time for survival predictions and for plotting the Kaplan-Meier curve. {phang} {opt ntimevar(#)} specifeis the number of timepoints to calculate the predicted survival curve at. The default is 100. {phang} {opt pitime(#)} is for use when the {cmd:stpm3} model has time-dependent effects and the user wants to use a prognostic index to define the groups. The prognostic index will be time-dependent, so this option specifies which time the prognostic index is calculated at. {bf: Additional options for {cmd:stpm3cif}} {phang} {opt compet1(#)} specifies the value of the event indicator for the first competing event. See {help stcompet} for more details. {phang} {opt competk(#)} specifies the value of the event indicator for the {it:k}th competing event. See {help stcompet} for more details. {phang} {opt crmodels(modellist)} specifies the competing risks models. {marker remarks}{...} {title:Remarks} Some remarks {marker examples}{...} {title:Examples} {pstd} All examples use the Roterdam data available on my website. First load and {cmd:stset} the data and then all examples are clickable. You will need to clear data in memory before running. {pmore} {stata "use https://pclambert.net/data/rott2b.dta":. use "https://pclambert.net/data/rott2b.dta"}{p_end} {pmore} {stata "stset os, failure(osi=1) scale(12) exit(time 120)":. stset os, failure(osi=1) scale(12) exit(time 120)}{p_end} {title:Example 1:} {pmore} Fit {cmd:stpm3} model {pmore} {stata "stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)":. stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)}{p_end} {pmore} Compare marginal predictions in 5 groups (the default) based on prognostic index {pmore} {stata "stpm3km":. stpm3km}{p_end} {pmore} Compare marginal predictions in 4 groups based on distribution of age {pmore} {stata "stpm3km age, groups(4)":. stpm3km age, groups(4)}{p_end} {pmore} Compare marginal predictions in based on factor variable {cmd:hormon} {pmore} {stata "stpm3km i.hormon":. stpm3km hormon, factor}{p_end} {title:Example 2: Competing risks} {pmore} Fit model with death due to cancer defined as the event. {pmore} {stata "stset os, failure(cause=1) scale(12) exit(time 120)":. stset os, failure(cause=1) scale(12) exit(time 120)}{p_end} {pmore} {stata "stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)":. stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)}{p_end} {pmore} {stata "estimates store cancer":. estimates store cancer}{p_end} {pmore} Fit model with death due to other causes defined as the event. {pmore} {stata "stset os, failure(cause=2) scale(12) exit(time 120)":. stset os, failure(cause=2) scale(12) exit(time 120)}{p_end} {pmore} {stata "stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)":. stpm3 @ns(age,df(3)) i.hormon enodes, scale(lncumhazard) df(4)}{p_end} {pmore} {stata "estimates store other":. estimates store cancer}{p_end} {pmore} Compare marginal predictions in 5 groups (the default) based on prognostic index (of cancer model) {pmore} {stata "stset os, failure(cause=1) scale(12) exit(time 120)":. stset os, failure(cause=1) scale(12) exit(time 120)}{p_end} {pmore} {stata "stpm3aj, crmodels(cancer other) compet1(2)":. stpm3aj, crmodels(cancer other) compet1(2)}{p_end} {pmore} Calibration in-the-large {pmore} {stata "stpm3aj, crmodels(cancer other) compet1(2) groups(1)":. stpm3aj, crmodels(cancer other) compet1(2) groups(1)}{p_end} {pmore} Compare marginal predictions in 4 groups based on distribution of age {pmore} {stata "stpm3aj age, crmodels(cancer other) compet1(2) groups(4)":. stpm3aj age, crmodels(cancer other) compet1(2) groups(4)}{p_end} {title:Author} {p 5 12 2}{bf:Paul C. Lambert}{p_end} {p 5 12 2}Cancer Registry of Norway{p_end} {p 5 12 2}National Institute of Public Health{p_end} {p 5 12 2}Oslo, Norway{p_end} {p 5 12 2}{it: and}{p_end} {p 5 12 2}Department of Medical Epidemiology and Biostatistics{p_end} {p 5 12 2}Karolinska Institutet{p_end} {p 5 12 2}Stockholm, Sweden{p_end} {p 5 12 2}pclt@kreftregisteret.no{p_end}