modelling.md
In TrialEmulation: Causal Analysis of Observational Time-to-Event Data

Modelling works with where_case

{
  "type": "list",
  "attributes": {
    "names": {
      "type": "character",
      "attributes": {},
      "value": ["names", "estimate", "robust_se", "2.5%", "97.5%", "z", "p_value"]
    },
    "row.names": {
      "type": "integer",
      "attributes": {},
      "value": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]
    },
    "class": {
      "type": "character",
      "attributes": {},
      "value": ["data.frame"]
    }
  },
  "value": [
    {
      "type": "character",
      "attributes": {},
      "value": ["(Intercept)", "assigned_treatment", "factor(followup_time)1", "factor(followup_time)2", "factor(followup_time)3", "factor(followup_time)4", "X1", "X2", "X3", "X4", "age_s"]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-5.7244369, -1.16602145, -0.06608159, -1.2440854, -15.48041975, -15.41173261, 2.12009182, -0.30716764, 0.55948724, 1.33660843, 0.08468846]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [1.30139124, 1.25619541, 0.54388987, 1.21696175, 0.78275918, 0.87168045, 0.865842, 0.44024115, 1.07397483, 0.97956078, 0.47265276]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-8.27516372, -3.62816445, -1.13210575, -3.62933043, -17.01462775, -17.1202263, 0.42304149, -1.17004031, -1.54550343, -0.58333071, -0.84171095]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.17371008, 1.29612155, 0.99994256, 1.14115963, -13.94621176, -13.70323892, 3.81714214, 0.55570502, 2.66447792, 3.25654756, 1.01108787]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-4.39870559, -0.92821662, -0.12149812, -1.022288, -19.7767336, -17.68048435, 2.44858971, -0.69772588, 0.52095005, 1.36449769, 0.1791769]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.00001089, 0.35329522, 0.90329651, 0.30664461, 0, 0, 0.01434167, 0.48534864, 0.60240157, 0.17241098, 0.8577988]
    }
  ]
}

trial_msm works with analysis_weights = unweighted

{
  "type": "list",
  "attributes": {
    "names": {
      "type": "character",
      "attributes": {},
      "value": ["names", "estimate", "robust_se", "2.5%", "97.5%", "z", "p_value"]
    },
    "row.names": {
      "type": "integer",
      "attributes": {},
      "value": [1, 2, 3, 4, 5, 6, 7, 8, 9]
    },
    "class": {
      "type": "character",
      "attributes": {},
      "value": ["data.frame"]
    }
  },
  "value": [
    {
      "type": "character",
      "attributes": {},
      "value": ["(Intercept)", "assigned_treatment", "trial_period", "followup_time", "X1", "X2", "X3", "X4", "age_s"]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.06458286, -1.87374327, -0.4264429, -0.51804601, 1.25103652, 0.29782921, 0.19461737, 1.23038439, 0.00277732]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.9149175, 1.26128915, 0.35243196, 0.34753752, 0.62776641, 0.50601482, 0.91360903, 0.87869213, 0.42536068]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-4.85782116, -4.34587001, -1.11720954, -1.19921955, 0.02061436, -0.69395983, -1.59605632, -0.49185219, -0.83092962]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-1.27134456, 0.59838347, 0.26432373, 0.16312754, 2.48145868, 1.28961825, 1.98529106, 2.95262096, 0.83648426]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.3495729, -1.48557789, -1.21000067, -1.49061892, 1.99283763, 0.58857805, 0.21302041, 1.40024514, 0.00652933]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.00080936, 0.13739081, 0.22627863, 0.13606157, 0.04627923, 0.55614436, 0.83131104, 0.16143992, 0.99479039]
    }
  ]
}

trial_msm works with analysis_weights = p99

{
  "type": "list",
  "attributes": {
    "names": {
      "type": "character",
      "attributes": {},
      "value": ["names", "estimate", "robust_se", "2.5%", "97.5%", "z", "p_value"]
    },
    "row.names": {
      "type": "integer",
      "attributes": {},
      "value": [1, 2, 3, 4, 5, 6, 7, 8, 9]
    },
    "class": {
      "type": "character",
      "attributes": {},
      "value": ["data.frame"]
    }
  },
  "value": [
    {
      "type": "character",
      "attributes": {},
      "value": ["(Intercept)", "assigned_treatment", "trial_period", "followup_time", "X1", "X2", "X3", "X4", "age_s"]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.46693136, -1.63471307, -0.39194898, -0.75152527, 1.31546169, 0.23796332, 0.54138231, 1.023308, -0.06542283]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [1.00800452, 1.26704723, 0.35471228, 0.37376728, 0.62056262, 0.5696091, 0.99111245, 0.80075479, 0.43668267]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-5.44262022, -4.11812564, -1.08718506, -1.48410914, 0.09915896, -0.87847052, -1.4011981, -0.54617138, -0.92132087]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-1.4912425, 0.8486995, 0.30328709, -0.0189414, 2.53176442, 1.35439716, 2.48396272, 2.59278739, 0.79047521]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.43940061, -1.29017532, -1.10497719, -2.010677, 2.11978881, 0.417766, 0.54623702, 1.27792929, -0.14981779]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.000583, 0.19698979, 0.26916947, 0.04435958, 0.03402386, 0.67611821, 0.58490302, 0.20127436, 0.88090837]
    }
  ]
}

trial_msm works with analysis_weights = weight_limits

{
  "type": "list",
  "attributes": {
    "names": {
      "type": "character",
      "attributes": {},
      "value": ["names", "estimate", "robust_se", "2.5%", "97.5%", "z", "p_value"]
    },
    "row.names": {
      "type": "integer",
      "attributes": {},
      "value": [1, 2, 3, 4, 5, 6, 7, 8, 9]
    },
    "class": {
      "type": "character",
      "attributes": {},
      "value": ["data.frame"]
    }
  },
  "value": [
    {
      "type": "character",
      "attributes": {},
      "value": ["(Intercept)", "assigned_treatment", "trial_period", "followup_time", "X1", "X2", "X3", "X4", "age_s"]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.4667822, -1.63456472, -0.39192411, -0.75276356, 1.31501668, 0.23791974, 0.54116841, 1.0243666, -0.06489607]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [1.00805417, 1.26721153, 0.3545763, 0.37577944, 0.62039488, 0.56993745, 0.99056982, 0.79963855, 0.43664379]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-5.44256838, -4.11829932, -1.08689366, -1.48929126, 0.09904272, -0.87915767, -1.40034843, -0.54292496, -0.92071789]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-1.49099601, 0.84916988, 0.30304544, -0.01623586, 2.53099064, 1.35499715, 2.48268526, 2.59165816, 0.79092575]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.43908322, -1.28989098, -1.10533082, -2.00320581, 2.11964464, 0.41744887, 0.54632031, 1.28103704, -0.14862475]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.00058369, 0.19708851, 0.26901627, 0.0451552, 0.03403603, 0.67635012, 0.58484577, 0.20018066, 0.88184974]
    }
  ]
}

trial_msm works with sample weights

{
  "type": "list",
  "attributes": {
    "names": {
      "type": "character",
      "attributes": {},
      "value": ["names", "estimate", "robust_se", "2.5%", "97.5%", "z", "p_value"]
    },
    "row.names": {
      "type": "integer",
      "attributes": {},
      "value": [1, 2, 3, 4, 5, 6, 7, 8, 9]
    },
    "class": {
      "type": "character",
      "attributes": {},
      "value": ["data.frame"]
    }
  },
  "value": [
    {
      "type": "character",
      "attributes": {},
      "value": ["(Intercept)", "assigned_treatment", "trial_period", "followup_time", "X1", "X2", "X3", "X4", "age_s"]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.86140786, -1.51756024, -0.37353376, -0.82824879, 1.41958257, 0.14505724, 0.72382671, 0.86103731, 0.05905786]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [1.08163206, 1.19667548, 0.43416259, 0.36449798, 0.72582293, 0.56898797, 1.0903126, 0.881077, 0.40424615]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-5.9814067, -3.86304417, -1.22449244, -1.54266483, -0.00303037, -0.97015917, -1.41318598, -0.86587361, -0.7332646]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-1.74140902, 0.82792369, 0.47742493, -0.11383275, 2.8421955, 1.26027366, 2.8608394, 2.58794822, 0.85138033]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [-3.5699828, -1.26814685, -0.86035454, -2.27230009, 1.95582493, 0.25493904, 0.66387081, 0.97725546, 0.14609382]
    },
    {
      "type": "double",
      "attributes": {},
      "value": [0.000357, 0.20474551, 0.38959364, 0.02306839, 0.05048578, 0.79877019, 0.506773, 0.3284427, 0.88384733]
    }
  ]
}

fit_msm works

Code
  fm_01
Output
  Trial Sequence Object 
  Estimand: Intention-to-treat

  Data: 
   - N: 725 observations from 89 patients 
          id period treatment    x1           x2    x3        x4   age      age_s
       <int>  <int>     <num> <num>        <num> <int>     <num> <num>      <num>
    1:     1      0         1     1  1.146148362     0 0.7342030    36 0.08333333
    2:     1      1         1     1  0.002200337     0 0.7342030    37 0.16666667
   ---                                                                           
  724:    99      6         1     1 -0.033762356     1 0.5752681    71 3.00000000
  725:    99      7         0     0 -1.340496520     1 0.5752681    72 3.08333333
       outcome censored eligible time_on_regime        wt       wtC
         <num>    <int>    <num>          <num>     <num>     <num>
    1:       0        0        1              0 1.0111849 1.0111849
    2:       0        0        0              1 0.9562510 0.9562510
   ---                                                             
  724:       0        0        0              1 0.9269537 0.9269537
  725:       1        0        0              2 0.9448756 0.9448756

  IPW for informative censoring: 
   - Numerator formula: 1 - censored ~ x3 
   - Denominator formula: 1 - censored ~ x1 + x2 + x3 
   - Numerator model is pooled across treatment arms. Denominator model is not pooled. 
   - Model fitter type: te_stats_glm_logit 
   - View weight model summaries with show_weight_models()

  Sequence of Trials Data: 
  - Chunk size: 500 
  - Censor at switch: FALSE 
  - First period: 0 | Last period: Inf

  A TE Datastore CSV object 
  N: 1558 observations 
  Periods: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 
  Path: 
  Columns: id, trial_period, followup_time, outcome, weight, treatment, x1, x2, x3, assigned_treatment

  Outcome model: 
  - Formula: outcome ~ assigned_treatment + x1 + x2 + followup_time + x3 
  - Treatment variable: assigned_treatment 
  - Adjustment variables: x1 x2 x3 
  - Model fitter type: te_stats_glm_logit

  Model Summary:

   term               estimate std.error statistic p.value conf.low conf.high
   (Intercept)        -6.2053  0.846     -7.337    2.2e-13 -7.863   -4.548   
   assigned_treatment  2.0802  0.603      3.451    5.6e-04  0.899    3.262   
   x1                 -0.5030  0.543     -0.926    3.5e-01 -1.567    0.561   
   x2                  0.1017  0.308      0.330    7.4e-01 -0.503    0.706   
   followup_time      -0.0029  0.046     -0.062    9.5e-01 -0.094    0.088   
   x3                  0.7231  0.713      1.014    3.1e-01 -0.675    2.121

   null.deviance df.null logLik AIC BIC deviance df.residual nobs
   148           1557    -74.8  162 194 136      1552        1558

  Outcome data 
  N: 1558 observations from 89 patients in 18 trial periods 
  Periods: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 
           id trial_period followup_time outcome    weight treatment    x1
        <int>        <int>         <int>   <int>     <num>     <int> <int>
     1:     1            0             0       0 1.0000000         1     1
     2:     1            0             1       0 0.9562510         1     1
    ---                                                                   
  1557:    54           17             1       0 0.9456885         0     0
  1558:    54           17             2       0 0.9373179         0     0
              x2    x3 assigned_treatment sample_weight         w
           <num> <int>              <int>         <num>     <num>
     1: 1.146148     0                  1             1 1.0000000
     2: 1.146148     0                  1             1 0.9562510
    ---                                                          
  1557: 1.846423     1                  1             1 0.9456885
  1558: 1.846423     1                  1             1 0.9373179