Data structure for a discrete-time event history analysis Jane E. Miller, PhD The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Overview • Structure of most survey data: One record per respondent • Discrete-time event history analysis requires separate records for each person-time unit at risk of the event • Review: How to create one record per spell • How to create one record per person-time unit – Components of the dependent variable – Fixed characteristics – Time varying characteristics The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Data preparation for an event history • Survey data often contains one record per respondent • Continuous-time event history data contain one record per spell • Discrete-time event history analysis requires one record per person-time unit within each spell – E.g., one record for each person-month at risk of divorce, within each spell at risk of divorce The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Source data from survey: 1 record per respondent ID Date of Date of Date of 1st 1st birth marriage divorce 1 2/1/52. 2 Date of Date of Date of Date of 2nd 2nd Date of Date 1st Date last 1st child's 2nd child's marriage divorce death observed observed Gender birth birth . . . . 7/15/85 10/1/10 F . . 7/15/69 6/22/10. . . . 9/21/85 11/5/10 M . . 1/1/97 10/1/04. . 10/8/85 5/1/05 M 12/5/95. 10/1/02 12/2/85 10/2/02 F 9/21/64 5/11/67 3 3/1/65 8/1/90 4 3/1/42 6/1/63. . . The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Example timelines for study of divorce M = Married D = Divorced L = Lost to follow-up O = Censored by end of study. X = Died Case 1: Never married -> no spells Case 2: Married once, censored by end of survey Case 3: Married twice, lost to follow-up before end of survey Case 4: Married once, died before end of survey M O M Not married -> not at risk of divorce -> not part of a spell M D M L X End of observation period The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Continuous-time event history data • One record for each period at risk (spell) – Duration of overall spell – Event indicator at end of spell Date Duration Status Divorce Age first Age at Age last # kids at Spell # spell of spell at end event observed start of observed start of ID (marriage #) started (mos.) of spell indicator (yrs) spell (yrs) (yrs) Gender spell 2 1 6/22/10 3.5 0 0 16 40 41 male 0 3 1 8/1/90 76.5 1 1 20 25 45 male 0 3 2 10/1/04 6.5 2 0 20 39 45 male 1 4 1 6/1/63 474.5 3 0 43 21 60 female 0 Event history analysis: discrete time data The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history timeline: Discrete time specification Four person-month units Case 2, Continuous time version: One four-month spell Married 6/22/2010 Last surveyed 11/5/2010 Case 2, Discrete-time version: Each person-month unit becomes one record -> unit of analysis. All records for each spell include respondent ID and other characteristics. 1st person-month Married O O O 3rd person-month O O 4th person-month O 2nd person-month O = Censored End of survey The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Discrete-time data set: ID codes on person-time records One record per spell ID 2 3 3 Duration Status Spell # of spell at end Divorce (marriage #) (mos.) of spell indicator 1 4 0 0 1 77 1 1 2 7 2 0 • Each person-month record carries the respondent ID • Each record within a given spell also includes the spell # for that respondent One record per person-month ID 2 2 2 2 3 3 3 3 3 3 3 3 3 Spell # Record # (marriage #) w/in spell 1 1 1 2 1 3 1 4 1 1 1 2 1 3 1 … 1 77 2 1 2 2 2 … 2 7 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Record number within spell One record per spell ID 2 3 3 Duration Status Spell # of spell at end Divorce (marriage #) (mos.) of spell indicator 1 4 0 0 1 77 1 1 2 7 2 0 • Each month in a spell will generate one person-month record, e.g., – respondent #2 is observed for 4 months -> 4 person-month records – respondent #3 contributes a total of 84 records • 77 in his first spell • 7 in his second spell One record per person-month ID 2 2 2 2 3 3 3 3 3 3 3 3 3 Spell # Record # (marriage #) w/in spell 1 1 1 2 1 3 1 4 1 1 1 2 1 3 1 … 1 77 2 1 2 2 2 … 2 7 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Month counter within spell One record per spell Duration Status Spell # of spell at end Divorce ID (marriage #) (mos.) of spell indicator 2 1 4 0 0 3 1 77 1 1 3 2 7 2 0 The “month # within spell” counter indicates the start time of the person-month at risk for that record. E.g., the first record for a given spell starts at baseline (time point 0). One record per person-month ID 2 2 2 2 3 3 3 3 3 3 3 3 3 month # Spell # Record # within (marriage #) w/in spell spell 1 1 0 1 2 1 1 3 2 1 4 3 1 1 0 1 2 1 1 3 3 1 … … 1 77 76 2 1 1 2 2 2 2 … … 2 7 6 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Duration measure for each record within spell One record per person-month One record per spell Duration Status Spell # of spell at end Divorce ID (marriage #) (mos.) of spell indicator 2 1 4 0 0 3 1 77 1 1 3 2 7 2 0 The duration measure will = 1 time units for all person-time records within a given spell EXCEPT = 0.5 for the last month in a spell ID 2 2 2 2 3 3 3 3 3 3 3 3 3 PersonSpell # Record month # months (marriage # w/in within w/in #) spell spell record 1 1 0 1 1 2 1 1 1 3 2 1 1 4 3 .5 1 1 0 1 1 2 1 1 1 3 3 1 1 … … 1 1 77 76 .5 2 1 1 1 2 2 2 1 2 … … 1 2 7 6 .5 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Status indicator for each record within spell One record per person-month One record per spell Duration Status Spell # of spell at end Divorce ID (marriage #) (mos.) of spell indicator 2 1 4 0 0 3 1 77 1 1 3 2 7 2 0 The indicator for status at end of record will = 0 for all person-time records within a given spell EXCEPT the last one because by definition they end in censoring (the spell is not yet complete) ID 2 2 2 2 3 3 3 3 3 3 3 3 3 Person- Status Spell # Record month # months at end (marriage # w/in within w/in of #) spell spell record record 1 1 0 1 0 1 2 1 1 0 1 3 2 1 0 1 4 3 .5 0 1 1 0 1 0 1 2 1 1 0 1 3 3 1 0 1 … … 1 0 1 77 76 .5 1 2 1 1 1 0 2 2 2 1 0 2 … … 1 0 2 7 6 .5 2 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Status indicator for last record within spell One record per spell One record per person-month Duration Status Spell # of spell at end Divorce ID (marriage #) (mos.) of spell indicator 2 1 4 0 0 3 1 77 1 1 3 2 7 2 0 The indicator for status at end of record for the last persontime record within each spell will take on the value of the status indicator for the overall spell ID 2 2 2 2 3 3 3 3 3 3 3 3 3 Person- Status Spell # Record month # months at end (marriage # w/in within w/in of #) spell spell record record 1 1 0 1 0 1 2 1 1 0 1 3 2 1 0 1 4 3 .5 0 1 1 0 1 0 1 2 1 1 0 1 3 3 1 0 1 … … 1 0 1 77 76 .5 1 2 1 1 1 0 2 2 2 1 0 2 … … 1 0 2 7 6 .5 2 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Event indicator for each record within spell One record per person-month One record per spell Duration Status Spell # of spell at end Divorce ID (marriage #) (mos.) of spell indicator 2 1 4 0 0 3 1 77 1 1 3 2 7 2 0 ID 2 2 2 2 3 3 3 3 3 3 3 3 3 month # Divorce Spell # Record # within indicator (marriage #) w/in spell spell for record 1 1 0 0 1 2 1 0 1 3 2 0 1 4 3 0 1 1 0 0 1 2 1 0 1 3 3 0 1 … … 0 1 77 76 1 2 1 1 0 2 2 2 0 2 … … 0 2 7 6 0 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Fixed covariates for each person-time record Age, number of children at start of spell, and gender do not change during the course of a spell, so they have the same value for each person-time record within a given spell ID 2 2 2 2 3 3 3 3 3 3 3 3 3 Spell # Record # (marriage #) w/in spell 1 1 1 2 1 3 1 4 1 1 1 2 1 3 1 … 1 77 2 1 2 2 2 … 2 7 Event history analysis: discrete time data month # Divorce Age at within indicator start of spell for record spell (yrs) Gender 0 0 40 male 1 0 40 male 2 0 40 male 3 0 40 male 0 0 25 male 1 0 25 male 3 0 25 male … 0 25 male 76 1 25 male 1 0 39 male 2 0 39 male … 0 39 male 6 0 39 male # children at start of spell 0 0 0 0 0 0 0 0 0 1 1 1 1 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Example timelines for number of children as time-varying covariate in study of divorce Columns reordered into chronological order ID Date of Date of 2nd Date of Date 1st Date of 1st 1st child's child's birth observed marriage birth birth 3 3/1/65 10/8/85 8/1/90 12/5/95. 4 3/1/42 12/2/85 6/1/63 9/21/64 5/11/67. Date of Date of Date of 1st 2nd 2nd divorce marriage divorce 1/1/97 10/1/04. . M C Date of death . . D Date last observed 5/1/05 10/1/02 M 10/2/02 L Case 3: No kids Case 4: M No kids C One kid One kid X C Two kids M = Married D = Divorced C = Child born L = Lost to follow-up O = Censored by end of study. X = Died The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Discrete time with time-varying covariates • Case 3 has his first child 64 months into his first marriage, and no additional children while observed. # kids at start of record is 0 for his first 63 records of spell 1 1 for records 64 through 77 of spell 1 1 for all records in spell 2 ID 3 3 3 3 3 3 3 3 Divorce # kids at month # indicator for start of Spell # w/in spell record spell 1 0 0 0 1 1 0 0 1 … 0 0 1 64 0 0 1 77 1 0 2 0 0 1 2 … 0 1 2 6 0 1 # kids at start of record 0 0 0 1 1 1 1 1 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Discrete time with time-varying covariates • Case 4 has her first child 15 months into her marriage, a second child in month 47 after marriage. For her the # kids at start of record is 0 for her first 15 records 1 for records 15 through 46 2 for records 47 or higher, all in spell 1 ID 4 4 4 4 4 4 4 Divorce # kids at month # indicator for start of Spell # w/in spell record spell 1 0 0 0 1 … 0 0 1 15 0 0 1 … 0 0 1 47 0 0 1 … 0 0 1 474 0 0 # kids at start of record 0 0 1 1 2 2 2 The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Presenting information on event history construction: Background work • Most of the gory details of creating an event history are part of behind-the-scenes work – Important to do consistency checks to make sure event histories were created correctly given • • • • Original data source of information for timeline construction Type of event under study Fixed covariates Time-varying covariates – E.g., correct • • • • Number of spells per respondent Number of person-time records for each spell Duration and event indicators for each person-time record Values of fixed- and time-varying covariates for each person-time record The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Presenting information on event history construction • In the data and methods section, describe: – Original data source of information for timeline construction • Dates, status, duration of events – – – – – Type of event under study Unit of person-time (e.g., person-years, person-months) What constitutes censoring Fixed covariates Time-varying covariates • Source(s) of information for determining timing of changes in those variables • See checklist in chapter 17 of Writing about Multivariate Analysis, 2nd Edition for more detail on what to report The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Summary • A discrete-time event history analysis requires a separate record for each person-time unit at risk of the event • For each respondent, create correct number of spells • For each spell, calculate – Correct number of person-time units – Components of the dependent variable • Duration measure • Event indicator – Fixed characteristics – Time-varying characteristics • In data and methods section, describe data sources and variables for the event history The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Suggested resources • Allison, P. D. 2010. Survival Analysis Using the SAS System: A Practical Guide, 2nd Edition. Cary, NC: SAS Institute. • Miller, J. E. 2013. The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. University of Chicago Press, chapter 17. The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Suggested online resources • Podcast on data structure for a continuoustime event history analysis The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Suggested exercises • Study guide to The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. – Question #3a in the problem set for chapter 17 – Suggested course extensions for chapter 17 • “Reviewing” exercises #2a through 2h • “Applying statistics and writing” exercises #1 and 2a The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data Contact information Jane E. Miller, PhD jmiller@ifh.rutgers.edu Online materials available at http://press.uchicago.edu/books/miller/multivariate/index.html The Chicago Guide to Writing about Multivariate Analysis, 2nd Edition. Event history analysis: discrete time data