2016paper - Adaptive Modelling of Trauma: Development and Recovery of Patients

Here you can find all material used for experiments and the link for the paper: http://dx.doi.org/10.1016/j.procs.2016.07.473

You are free for use as the base for new researchers or improve it, since reference this work.


ICPSR_09329: real data used in experiments.

  • You can find the rouge data and a worksheet with processed data. In that worksheet you can see how the variables was counted for each input/output of the model.
  • Data.xlsx ds1, ds2, ds3, ... are the sheets with rouge data.
  • Each column represent one variable, the used variables are identified by its name relative to description in: 09329-Codebook.pdf
  • trauma, trigger, social support, ... are the compilation of variables used. Each line is a register (different real case/person).
  • You can see how variables were counted by observing the formulas inside cells. Each variable is composed by 3 columns. Example: CUTOFF | weight | var value where:
    • CUTOFF is the variable and its rouge value
    • Weight is the importance of that variable in the sum of values for the input of the model, for this case: SOCIAL SUPPORT
    • var value is the final value of this variable for the input of the model.
    • at right of each sheet you find one or more of this tags: FI-N FI-P SI-N SI-P, they means: FI-N:
    • This columns aggregate the sum of values of variables and are the inputs of the model.
  • We just used the second interview value in our experiments
  • We only used complete registers without missing data, that is found in: data_resume.xlsx
  • Other files in this folder correspond to original database. See link-to-database

matlab_v160202: last matlab version.

All files named t_* is perform a different test. Other files are auxiliary files used by the tests.
     data.m have the data imported from excel file.

Material and Methods

Here is described a brief overview of the variables used from real data and how the system used them.

A real data set was applied to the model to validate it. The data set collection consists of questionnaires answered by victims some time after a trauma occurred. The questions provide information about the trauma, feelings and social aspects of victims [Cook et al., 1990], it reflect the life aspects of patients from 2 to 8 months after trauma.  Some of those answers were translated to inputs and outputs of the model, and one scenario for each individual was constructed. The answers of the computational model were compared with the answers from the real dataset. In total 257 scenarios were constructed, one for each sample of the data set. Table 2 shows the data used for each sample connected with inputs and outputs of the model. For each question, there are different levels of answers, vary from 1-5 characterizing the intensity of each measured aspect. Groups of variables were used to define each input and output of the model, the sum of  answers for each group of variables were normalized to values between 0 and 1, where 0 correspond to low intensity and 1 high intensity. Table 2 shows the groups of variables related to each input or output of the system. Details about the each question and its answers are available in [Cook et al., 1990].

The internal states and outputs of the model have to fit perfect with the constraints related in psychological literature and with the real data collected, along the time simulation. Each sample has specific tuning of weights in the model. That also reflects the personality of each patient. The values of states start with random values. After the first simulations steps, those states assume values coherence with the tuned personality and the external stimulus. Some of the internal weights are more important to define the personality and how people respond to events, while others adjust specific characteristics.

After tuning the model for each person is easy to identify groups of personalities with slight variations and be sure that the model could reproduce real patterns for different types of trauma in different intensities and post-traumatic conditions those mapped personalities. With those groups of people modeled, new experiments was conducted connecting different or similar patients in groups. The connections are performed linking esb of patients with the inputs wsss and wsnc of the others in the group.


Variables selected from data set

Positive Social Support

FFAMHELP: since the last interview did family help since the incident

FFRIHELP: since the last interview did friends help since the incident

Negative Contagion

FCUTOFF: since the last interview have cutted yourself off from friends

FCUTOUT: since the last interview have cutted yourself off from places used to go


CHARGE1 to CHARGE3: type trauma, until 3 traumas by event

INJURIES1 to INJURIES4: Did you receive any injuries


FAFRAID1 to FAFRAID12: are you disturbed by violent movies, parking lots, left car alone, knifes, strangers, etc.

FDIFFDAILY: are you having difficulties going about your usual daily activities


FPASTFIN: since last questionnaire how the financial things going on

FPASTPHYS: since last questionnaire, how you physically have you felt

FPASTSOC: since last questionnaire how the social life is going on

Emotional Response

FPASTEMO: Emotionally, how thing is going on


FTIME: days between First and Second questionnaire

DAYS: days between incident and First questionnaire

Table 2. Relation data set variables used on input/output of the model.

[Cook et al., 1990] R. Cook, B. Smith and  A. Harrell. Helping Crime Victims: Levels of Trauma and Effectiveness of Services in Arizona, 1983-1984, Washington, DC: Georgetown University, Institute for Social Analysis [producer], 1984. Ann Arbor, MI: Inter-university Consortium for Political and SocialResearch [distributor], 1990 http://doi.org/10.3886/ICPSR09329.v1.

Daniel Formolo,
Jul 26, 2016, 7:44 AM
Daniel Formolo,
Jul 20, 2017, 10:21 AM