黑料网

Development of an Algorithm for Calculating the Risk of Terrorist-CBRN

Bolduc DL^1*, Marr J², King J³ and Dudley R⁴

¹Biodefense Program, Department of Public and International Affairs, George Mason University, USA

²Department of Computational and Data Sciences, George Mason University, USA

³Department of Mathematics, New Hampshire Community Technical College, USA

⁴Department of Political Science, George Mason University, USA

*Corresponding Author:

David L. Bolduc
Scientific Research Department
Armed Forces Radiobiology Research Institute (AFRRI)
Uniformed Services University of the Health Sciences
Bethesda, MD 20889
E-mail: david.bolduc@usuhs.edu

Received Date: April 28, 2012; Accepted Date: August 21, 2012; Published Date: August 24, 2012

Citation: Bolduc DL, Marr J, King J and Dudley R (2012) Development of an Algorithm for Calculating the ‘Risk’ of Terrorist-CBRN. J Bioterr Biodef 3:117. doi: 10.4172/2157-2526.1000117

Copyright: © 2012 Bolduc DL, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited..

Visit for more related articles at Journal of Bioterrorism & Biodefense

Abstract

In order to avert a disaster from a terrorist chemical, biological, radiological or nuclear (CBRN) attack, it is important to study the likelihood of terrorists using CBRN weapons. This study reports on the development of an algorithm for calculating the ‘risk’ of a terrorist seeking CBRN weaponry with 67.3 percent prediction accuracy. The algorithm was developed through four phases, Phase I proposed independent variables likely associated with Terrorist-CBRN (T-CBRN) derived from our interpretation of the literature; Phase II involved constructing a ‘Random Nations Matrix’ from 74 countries or locations of the world selected at random, for correlating the proposed independent variables; Phase III entailed the construction of a multivariate model from the independent variables which met our correlation criteria with T-CBRN; and finally in Phase IV, an algorithm was derived from the model design for calculating the risk of a terrorist seeking, acquiring and or using a CB

Introduction

As technology advances, so does the potential threat of terrorists using chemical, biological, radiological or nuclear (CBRN) weaponry. During the past decade, the horrific premise of the use of CBRN weapons commonly referred to as ‘Weapons of Mass Destruction’ (WMD) has continued to gain notoriety among U.S. policymakers, news media, and the academic analytical community. New developments leading to the growing concern over terrorist use of CBRN weapons have included the expansion of CBRN related technologies, materials, technical knowledge and the appearance of trends in terrorist tactics. These developments may suggest an increasing appreciation of attacks for inflicting mass-casualties (of which CBRN weapons would be an excellent means). The discovered interest in CBRN weapon technology by the terrorist group Al-Qaeda (Osama bin Laden), the actual biological and chemical terror attacks by the Rajneesh with Salmonella in 1984 and Aum Shinrikyo with Bacillus anthracis in 1993 (though failed) and sarin gas in 1995, support this possibility [1]. The 1995 sarin gas attacks by the Doomsday cult Aum Shinrikyo resulted in 12 deaths and approximately 6,000 injuries. This act, intended for inflicting mass casualties, confirmed to the world that this dreaded scenario was no longer a speculated supposition, but rather a frightening reality. In April 1995, a lone perpetrator ascribing to an extreme Right-ideology detonated a truck bomb destroying the Alfred P. Murrah Federal building in Oklahoma City. This incident, though technically not caused from a CBRN weapon, resulted in the deaths of 168 people and the wounding of more than 700. These and other key events marked the beginning of the ‘New Age of Terrorism.’ The New Age of Terrorism is characterized by pronounced changes in terrorist tactics, goals and motivations. The political motivations that were characteristic of the previous terrorist age of the 1980s and 1990s have now been replaced by religious inspirations. No longer are terrorists bound by societal norms that once prevented them from using horrific weaponry for inflicting high numbers of casualties. Mass casualties became the goal. The late Osama bin Laden, former leader of the Al-Qaeda international terrorist organization, is on record as overtly stating his interest in obtaining WMDs for engaging in attacks against America. During this same timeframe, the dissemination of dual-use technologies and materials needed for developing WMDs became widely available to countries throughout the world, including potential state-sponsors of terrorism. As WMD technology continued to expand throughout the world, controls on its technologies and materials continued to erode as the remains of the former Soviet Union continued to disintegrate [2,3].

U.S. policy makers examining the threat of terrorist-CBRN (T-CBRN) weapon capability were warned by the FBI’s National Infrastructure Protection Center (NIPC) that Al-Qaeda and affiliated groups were continuing ‘to enhance their capabilities to conduct effective mass-casualty chemical, biological, radiological, and nuclear attacks.’ NIPC further contended that Al-Qaeda was currently in possession of at the least, a ‘crude’ CBRN-weapon ‘capability’ [4]. Policymakers today find themselves in the difficult position of having to discern accurate T-CBRN weapon assessments from various intelligence suppositions. Driven by inherent doubts about the quality and accuracy of U.S. intelligence reports, coupled with the fear of the terrible premise that preludes T-CBRN, many U.S. policy makers have taken no chances in the drafting of policies estimating the T-CBRN threat. The lack of the capability to perform clear, concise T-CBRN analysis, has compelled many policy makers to hedge against unknown factors by adopting safer positions and ascribing to likely exaggerated assessments of T-CBRN capability [5].

A major problem with T-CBRN assessment is the lack of historical cases on the subject. There is only a relatively small data-set available for analyzing T-CBRN risk. Adding further to the difficulty in performing accurate analyses is the sketchiness of the details that do exist. Fortunately, however, this lack of information also means that the events involving the use of CBRN weapons by radical-perpetrators are relatively rare [5].

Over the past ten years, considerable scholarship has been applied to better define and typify the phenomena of T-CBRN [6]. A variety of cultural and political environments have been examined in an attempt to predict the likelihood of terrorists using CBRN weaponry. Various factors influencing terrorist’s use of CBRN have been proposed such technological development, embeddedness in the world economy, etc. [7]. Asal and Rethemeyer have suggested that cultural integration into the Westernized world could potentially give terrorists an enhanced awareness of what might be achievable in terms of CBRN-terrorism, but concede that their findings were negatively correlated when attempting to measure their integration variable [7]. Unfortunately, the majority of the scholarship is based on hypotheses derived from secondary literature on terrorism events involving CBRN weapons, or subjective evidence taken from a few high profile terrorist cases such as the 1995 Tokyo subway sarin gas attacks by the Japanese cult Aum Shinrikyo and the 2001 U.S. anthrax letter attacks by alleged U.S. researcher, Dr. Bruce Edwards Ivins. This means of analysis has led to considerable conjecture with an extremely wide range of T-CBRN threat assessments and speculations on possible perpetrators and motives. A new analytical methodology is needed to investigate this obscure area.

We believe that quantitative, analytical methodologies may be a better tactic for determining causal indicators of radical perpetrators that would seek, acquire and/or use CBRN weapons. The intent of this study was therefore to construct a mathematical matrix for correlating variables for determining indicators of non-state actors likely to seek, acquire and /or use a CBRN weapon. If we were successful in finding variables that correlated highly with T-CBRN events, our next phase would be in the placing of these indicators (correlating variables) in sequence to produce a statistical model for theoretically predicting the amount of T-CBRN activity.

The purpose of this research was to develop an algorithm for calculating the risk of terrorists using CBRN weapons. The study addressed two main objectives: to calculate the risk of CBRN weapons being used by a specific terrorist, and the risk of CBRN weapons being used by nonspecific terrorists in a specific country. Risk was defined as the risk of CBRN by a specific terrorist based on the variable values used in the calculation, or the risk of CBRN occurring by nonspecific terrorists in a specific country relative to the variable values used in the calculation.

A mathematical matrix was compiled and linked with the statistical program (ProStat Version 5.5 Poly Software International). Variables were entered into an algorithm template which was deduced from multivariate analysis. Depending on the question posed, the algorithm would derive either a number representing the degree of risk of a given terrorist electing to pursue CBRN, or the risk of CBRN events occurring within a specific country by nonspecific terrorists.

The study consisted of four phases: Phase I involved searching open literature and proposing independent variables derived from our interpretation of the literature that we believed were associated with T-CBRN; Phase II entailed constructing a Random Nations matrix (representing the total T-CBRN universe) for correlating variables suspected of being associated with T-CBRN. Phase III involved the construction of a multivariate model from the variables which met our correlation criteria with T-CBRN. Phase IV, our last phase, entailed the construction of an algorithm from the model design for calculating the risk of a given terrorist electing to pursue CBRN, or the risk of CBRN events occurring within a specific country by nonspecific terrorists. The pursuit of CBRN was defined as ‘the act of seeking, acquiring and/or using CBRN’.

Potential variables proposed from the open literature and formulation of hypotheses

Based on the literature, 38 independent variables that we believed could be significant indicators of T-CBRN were proposed (Table 1). The selection of these variables was driven by the availability of data and by our desire to select a broad range of potential terrorist-elements. Four categories were derived for organizing the variables by their subject matter:

Table 1: The 38 Independent Variables (IV) used in the construction of the ‘Random Nation’s Matrix’. The variables include the ‘Country of Interest’ and the ‘Country Where the Terrorist of Interest is Operating.

1. Technological, which included countries’ electricity production and state-CBRN technologies;

2. Population, which included countries’ economic, civil and cultural variables;

3. Religion, which included the percentage of major religions in the selected country;

4. Terrorist Elements, which included technical expertise, terrorist-ideologies, degree of terrorist violence etc.

These variables were correlated for significance with our dependent variable ‘T-CBRN’ in order to determine their relation with T-CBRN.

The dependent variable: T-CBRN

The single dependent variable used in the study was Terrorist- CBRN. The definition of this variable is ‘the total of all documented substate actor CBRN events in the country of interest.’ ‘Events’ was defined as ‘acts of seeking, acquiring and/or using a CBRN.’ All independent variables were correlated with this variable for determining their relation with T-CBRN incidents.

Terrorist country of origin

The definition of the terrorist’s ‘country of origin’ was the country where the terrorist group originated from. Although some terrorist groups may be based in many countries, the group has only one country from which it initially began which cannot change. Whether a group chooses to leave its country of origin or is forced out has no bearing on the classification. An example of this is the Al-Qaeda (Central) organization, which was forced out of Afghanistan by U.S. and allied forces. Wherever the Al-Qaeda Central group may now reside, we presume Afghanistan as its country of origin, as it was the country where the terrorist group initially developed [2].

Formulation of the twelve hypotheses

From the 38 variables deduced from the open literature, we formulated twelve non repetitive hypotheses representing theoretical categories of terrorism: Technical (electricity production, state resources devoted to CBRN), Population (political violence, presence of U.S. troops, Western influence), Religious (denomination of religion) and Terrorist Elements (CBRN-technical expertise, number of groups, group ideology, state sponsor). CBRN weapon technology is complex requiring a technical infrastructure, and various types of special equipment are therefore needed, such as centrifuges, incubators, fermenters, refrigerators/freezers etc. [15]. In order to operate this equipment, an electrical power source is required which would likely restrict CBRN weapon manufacturing activity to areas having sufficient electricity production. From this summation, we formulated our first hypothesis: Countries with high electricity production correlate highly with T-CBRN.

Based on the open literature, we surmised that many of the terrorists that had used CBRN weapons were based in countries that had at one time actually used CBRN weapons themselves [2]. We speculated that a possible explanation could be the availability of technical resources in these countries, such as the accessibility of trained personnel in the needed technologies, or in the access to necessary chemicals, organisms and equipment. We further surmised that there could even be deeper intrinsic ideological factors involved, such as a culturally inherent appreciation drawn from the knowledge of past state successes in deploying CBRN weapons. This observation led to the formation of our second hypothesis: Countries with a history of high resources devoted to state-CBRN correlate highly with T-CBRN.

We supposed that the attitudes of citizens in countries where political violence is a relatively common cultural occurrence will be influenced. It was our assumption that citizens may develop an intrinsic tolerance for highly violent acts committed by perpetrators and rogue government officials. Moreover we propose that terrorists groups coming from highly violent cultures may have a greater tolerance for violent acts, as well as an appreciation for violence. This violent cultural behavior may serve as a model of which to base their tactics. Post states that terrorists most likely to seek CBRN weapons are those that wish to exact high numbers of casualties [16]. We suspected that a high ‘politically violent’ environment may be an indicator of a terrorist group’s propensity in inflicting high incidences of violence and potentially an abnormally higher likelihood in choosing to use CBRN weapons. From this summation, we formulated our third and fourth hypotheses: Countries with a history of high political violence correlate highly with T-CBRN; Countries with a history of high political violence per capita correlate highly with T-CBRN.

The amount of CBRN-technical experts in a terrorist group may be an indicator as to whether the group may choose to pursue CBRN weapons. This was clearly the case with Aum Shinrikyo. This notorious Millennial-cult had approximately 800 technical experts [17]. Technical people with the appropriate scientific knowledge, skills and backgrounds may have a bias for developing CBRN weapons in the areas in which they are very fluent. Scientific, technically orientated members may suggest a group with a higher propensity for using CBRN weapons. From this summation we formulated our fifth hypothesis: Countries with terrorist groups that have members that are CBRN-technical experts correlate highly with T-CBRN.

A great deal of open-literature suggests that groups, such as Religious Fundamentalist that aspire to inflict mass casualties are likely to pursue CBRN weapons [2,16,17]. Post reports that in addition to Religious Fundamentalist groups, groups ascribing to Right-Wing extremism ideologies are also likely pursuers of CBRN weaponry, though may have difficulty in obtaining the capability [16]. From this literature, we formulated our sixth and seventh hypotheses: Countries with CBRNterrorists aspiring to Religious Fundamentalist ideologies correlate highly with T-CBRN; Countries with CBRN-terrorist groups aspiring to Right-Wing ideologies correlate highly with T-CBRN.

Asal and Rethemeyer surmised that cultural integration into the Westernized world may give non-state actors a stronger awareness of what might be achievable. They argue that the more embedded an organization’s host country is in Westernized culture the more likely it was for the organization to pursue CBRN weaponry capability. This notion, Asal and Rethemeyer concede was difficult to measure. They operationalized their concept by measuring the number of McDonald’s restaurants in a country. Their assumption was that the McDonald’s restaurants were a fitting representation of the Western culture and proceeded on the premise that by measuring the number of restaurants in a country, they would be able to discern ‘the depth of a country’s link to a larger world culture’ [7]. From Asal and Rethemeyer’s assumption, we formulated our eighth and ninth hypotheses: Countries which are highly embedded in Westernized culture (measured by the number of McDonald’s restaurants in the country of interest/or in the terrorist’s country of origin), correlate highly with T-CBRN; Countries which are highly embedded in the Westernized culture (measured by McDonald’s restaurants per capita in the country of interest/ or the terrorist’s country of origin), correlate highly with T-CBRN.

The number of U.S. troops occupying a terrorist’s country may be an indicator as to the degree of resentment a terrorist may have towards the United States. Over the past ten years, we have witnessed great increases in the events of terrorism resulting from the escalation of U.S. troops in Iraq and Afghanistan [2]. During the Invasion of Kuwait by Iraqi forces in 1990, the country of Saudi Arabia and its ruling House of Saud were suddenly at great risk from attacks. The massive army of Saddam Hussein had positioned itself in Kuwait within easy striking distance of Saudi’s most valuable oil fields (Hama). During this time, Saddam Hussein broadcasted a global appeal for the establishment of a Pan-Arab Islamism, calling for all loyal Muslims to rebel against their corrupt governments. In addition to the Royal Family’s persistent fear of civil war, the House of Saud was suddenly faced with the threat of Iraqi invasion. Although Saudi Arabia’s forces were well equipped and well trained, the sheer number of Saddam’s army was far too great for them. Bin Laden, realizing the great threat facing his beloved country, offered the services of his Mujahedeen army to King Fahd for protection. Fahd refused bin Laden’s offer, instead opting to allow the United States’ forces and its allies to deploy on Saudi Arabian lands. Bin Laden was greatly enraged by Fahd’s decline of his patriotic offer and even more so by the presence of foreign troops in his country which held the two holy cities of Mecca and Medina, and which he therefore viewed as a sacrilege of the highest order. After publicly voicing his contempt for King Fahd, bin Laden was quickly pressured to leave his country and so exiled himself to Khartoum in Sudan [18-21]. Throughout the remainder of the 1990s to the present, we have seen an increase in acts of terrorism from Al-Qaeda and groups aspiring to them in protest to American occupation [2].

From the historical events, we discerned that an occupation of U.S. troops will increase the animosity of terrorists towards the U.S./ West. We surmise that with the increases in troop occupation, greater animosity will develop, which in turn will result in a desire to use ‘high impact’ CBRN weapons for achieving resistance. From this assumpton, we formulated our tenth and eleventh hypotheses: Foreign countries with a high number of U.S. troops present correlate highly with T-CBRN; Foreign countries with a high number of U.S. troops present per capita correlate highly with T-CBRN.

The ramifications of states sponsoring terrorist groups could theoretically be considerable. A state sponsoring a terrorist group could, for instance, potentially enable it with CBRN weapon capability. It has also been suggested that states commissioning terrorists to deploy CBRN on its enemies, could theoretically create ‘plausible deniability’ thereby avoiding a retaliatory attack. Such deniability is theoretically possible with CBRN-terrorism because it is easily misidentified [22]. From the preceding supposition, we formulated our twelfth hypothesis: Countries that sponsor terrorist groups correlate highly with T-CBRN.

Methods and Parameters

All data regarding sub-state actor CBRN activity was taken from the Monterey WMD Terrorism Database. Data that occurred from ‘unknown’ perpetrators or was classified as a ‘Hoax/Prank’ was excluded [2].

Definition of a CBRN event

For the study we defined a ‘CBRN Event’ as ‘the act of seeking, acquiring and/or using a CBRN weapon.’ Every documented independent incident in the Monterey WMD Terrorism Database of a perpetrator seeking, acquiring and/or using a CBRN weapon counted as an ‘event’ [2].

The random nations matrix

A ‘Random Nations Matrix’ was constructed and merged with statistical software (ProStat Version 5.5 Poly Software International, Pearl River, NY 10965). This matrix consisted of 74 nations and/or world locations selected at random from the CIA World Fact Book (2009-2010) (Table 2) [8]. We selected n = 74 in order to be well within the range of the normalized ‘Z’ distribution [24]. The 38 variables deduced from the literature were tested for their correlative values with the T-CBRN variable and with each other. Significance levels were set at p< 5.0 x 10^-2 and all associations were graphed and checked for linearity using the ProStat software [24].

Table 2: The 74 nations and/or locations of the world that were selected at random and used to construct the ‘Random Nation’s Matrix’.

The T-CBRN variable

The only dependent variable used in our study was the T-CBRN variable. All 38 independent variables were correlated with this variable for determining their relation with T-CBRN incidences.

Testing the 38 Independent Variables Derived from Our Interpretation of the Literature, Against the T-CBRN Variable

Least squares regression analysis was performed on the 39 variables (38 independent, 1 dependent, T-CBRN) on the matrix. The variables were tested for their correlative values with the T-CBRN variable and with each other. Significance levels were set at p< 0.05 and all associations were graphed and checked for linearity using the ProStat software.

Constructing the multivariate model

Many of the 38 independent variables were intercorrelated with each other and we were therefore only able to select two of the variables (Political Violence Per Capita and T-CBRN Technical Experts) for construction of the multivariate model. When choosing independent variables for multivariate modeling, it is vital to test for correlation between them to ensure that the independent variables are not intercorrelated. By using non-intercorrelated independent variables, one is able to achieve full predictive power from each of them for the dependent variable.

The predictive model was comprised of two independent variables (all that were discerned from the regression analysis) and the dependent variable. This model was used for deriving an algorithm for calculating risk of CBRN activity of specific terrorists and T-CBRN activity occurring in specific countries.

Designation of the variables used for constructing the multivariate model

X1 = Correlate 1(Independent Variable 1)

X2 = Correlate 2 (Independent Variable 2)

α = Alpha Coefficient (Y intercept)

b1 = Beta Coefficient 1 (The amount of change 1 unit of X1 produces in Y)

b2 = Beta Coefficient 2 (The amount of change 1 unit of X2 produces in Y)

Y= Dependent Variable (Total Number of Lone Terrorists/Terrorist Groups that Seek, Acquire and/or Use CBRN in the Country of Interest)

Mathematically written the multivariate model appears as: Y= α + (b1) (X1) + (b2) (X2) + error

Results

Testing the twelve hypotheses on the random nations matrix

Our 12 hypotheses derived from the variables discerned from the available literature, were tested on our Random Nations Matrix, the following correlation values were derived:

Testing of Hypothesis #1: Countries with high electricity production correlate highly with T-CBRN

Electricity Production with T-CBRN, (r) = 0.2126, p< 0.069, n= 74. In evaluating our first hypothesis, we detected no correlation between Electricity Production and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #2: Countries with a history of high resources devoted to state-CBRN correlate highly with T-CBRN

State-CBRN with T-CBRN, (r)=0.6891, p≅0.00001, n=74. In evaluating our second hypothesis, we detected a high correlation between State-CBRN and T-CBRN and the hypothesis was accepted.

Testing of Hypothesis #3: Countries with a history of high political violence correlate highly with T-CBRN

Political Violence with T-CBRN, (r) = 0.0761, p < 0.5192, n = 74. In evaluating our third hypothesis, we detected a low correlation between Political Violence and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #4: Countries with a history of high political violence per capita correlate highly with T-CBRN

Political Violence Per Capita with T-CBRN, (r) = 0.4542, p < 0.0004, n = 74. In evaluating our fourth hypothesis, we detected a high correlation between Political Violence Per Capita and T-CBRN and the hypothesis was accepted.

Testing of Hypothesis #5: Countries with terrorist groups that have members that are CBRN-technical experts correlate highly with T-CBRN

CBRN-Technical Experts with T-CBRN, (r) = 0.7443, p≅ 0.00001, n = 74. In evaluating our fifth hypothesis, we detected a high correlation between T-CBRN Technical Experts and T-CBRN and the hypothesis was accepted.

Testing of Hypothesis #6: Countries with CBRN-terrorists aspiring to ‘Religious Fundamentalism’ ideologies correlate highly with T-CBRN

Religious Fundamentalism Ideology with T-CBRN, (r) = 0.7883, p≅ 0.00001, n = 74. In evaluating our sixth hypothesis, we detected a high correlation between Religious Fundamentalism Ideology and T-CBRN and the hypothesis was accepted.

Testing of Hypothesis #7: Countries with CBRN-terrorist groups aspiring to ‘Right-Wing’ ideologies correlate highly with T-CBRN

Right-Wing Ideology with T-CBRN, (r) = 0.1284, p < 0.2754, n = 74. In evaluating our seventh hypothesis, no correlation was detected between Right-Wing Ideology and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #8: Countries which are highly embedded in the Westernized culture (measured by the number of McDonald’s restaurants in the country of interest/ or in the terrorist’s country of origin), correlate highly with T-CBRN

Number of McDonald’s Restaurants with T-CBRN, (r) = 0.1404, p < 0.2327, n = 74. In evaluating our eighth hypotheses, no correlation was detected between the Number of McDonald’s Restaurants and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #9: Countries which are highly embedded in the Westernized culture (measured by McDonald’s restaurants per capita in the country of interest/ or in the terrorist’s country of origin), correlate highly with T-CBRN

Number of McDonald’s Restaurants Per Capita with T-CBRN, (r) = -0.0228, p < 0.8469, n = 74. In evaluating our ninth hypothesis, we detected no correlation between the Number of McDonald’s Restaurants Per Capita and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #10: Foreign countries with a high number of U.S. troops present correlate highly with T-CBRN

Number of U.S. Troops with T-CBRN, (r) = 0.2340, p < 0.04448, n = 74. In evaluating our tenth hypothesis a weak correlation was detected between, the Number of U.S. Troops and T-CBRN and the hypothesis was accepted.

Testing of Hypothesis #11: Foreign countries with a high number of U.S. troops present per capita correlate highly with T-CBRN

Number of U.S. Troops Per Capita with T-CBRN (r) = -0.0273, p < 0.08176, n = 74. In evaluating our eleventh hypothesis, no correlation was detected between the Number of U.S. Troops Per Capita and T-CBRN and the hypothesis was rejected.

Testing of Hypothesis #12: Countries that sponsor terrorist groups correlate highly with T-CBRN

State-Sponsor Potential with T-CBRN, (r) = 0.6489, p≅ 0.00001, n = 74. In evaluating our twelfth hypothesis, a high correlation was detected between State-Sponsor Potential and T-CBRN and the hypothesis was accepted.

Multivariate model results Using the correlation data derived from our Random Nations matrix, we deduced that the best fitting independent variables for building our multivariate model were: Political Violence Per Capita at (r = 0.4502, p < 0.0004, n = 74), and Terrorist CBRN-Technical Experts at (r = 0.7443, p≅ 0.00001, n = 74).

To ensure that Political Violence Per Capita and Terrorist CBRNTechnical Experts were not essentially the same variable, correlation testing between the 3 variables was performed. No correlation was found between the 2 independent variables Political Violence Per Capita and Terrorist CBRN-Technical Experts, (r) = 0.1520, p < 0.1960, n = 74.

Quantifying the two independent variables

Quantification of the ‘Political Violence Per Capita’ variable was derived from the public database provided by the Center for Systemic Peace (CSP): ‘Major Episodes of Political Violence 1946-2009’ [11]. The CSP database tallied fatalities caused by major episodes of political violence from 1946-2009 throughout the world. Political violence was defined as events involving ‘international, civil, ethnic, communal, and genocidal violence and warfare’ [11]. ‘Per capita’ was calculated by dividing the total number of deaths in the country by the population of the country followed by multiplying the answer by 1,000.00 to equal the number of deaths per thousand.

The ‘Terrorist CBRN-Technical Experts’ variable was quantified by deriving the actual number of technical experts reported on the nonpublic database provided by the James Martin Center for Nonproliferation Studies (CNS) [1,2]. When quantifying technical experts in groups that were known to be associated with the terrorist group Al-Qaeda, the Al-Qaeda scale of ‘5’ was used since Al-Qaeda at that time (2009) was known to have at least five CBRN technical experts in their membership [1,2].

Combining of the independent variables

Once the two best fitting correlating independent variables were determined, the software derived beta coefficients, and the independent variables were merged to form a multivariate model.

Y = α + X1 + X2 + error

α = Y Intercept (Alpha Coefficient)

X1 = Correlate 1(Independent Variable 1)

X2 = Correlate 2 (Independent Variable 2)

Y= The Dependent Variable: Total Number of Lone Terrorists/ Terrorist Groups that Seek, Acquire and/or Use CBRN in the Country of Interest

Mathematically written the multivariate model appears as:

Y = α + (beta coefficient for X1) (X1) + (beta coefficient for X2) (X2) + error

The following multivariate model was constructed: T-CBRN (Total Number of Lone Terrorists/Terrorist Groups that Seek, Acquire and/ or Use CBRN) = -0.4189 + (0.146979864) (Degree of Political Violence Per Capita) + (1.61978789) (Number of Terrorist CBRN-Technical Experts) + error

(R) = 0.8204, p≅ 0.00001, n = 74, The square of R = 0.67304 = 67.30 %, F = 73.0785

67.30 % of the lone terrorists/terrorist groups that seek, acquire and/or use CBRN are explained/predicted by this model. Validating the multivariate model

Three tests were performed to validate our multivariate model (F test, Multiple R equation and three dimensional graphing). The F statistic if large, indicates that we have reached a value of p < 0.05. Our F value was at 73.07 indicating a p value of p≅ 0.00001. The Multiple R equation is a function of the correlations between the independent variables and the dependent variable, and also a function of the correlations of the independent variables among themselves (which must be close to zero thereby demonstrating ‘independence’). All data were graphed to detect patterns among the variables. This was done in order to determine whether nonlinear associations existed [23,24].

Running the multivariate model

In running the multivariate model, the independent variables are entered into the model:

Y = α + (b1) (X1) + (b2) (X2) + error

α = Y Intercept (Alpha Coefficient): -0.4189

Beta Coefficient for X1=0.1469

Beta Coefficient for X2=1.6197

X1=Political Violence Per Capita in the Terrorist Origin Country

X2 = Number of Terrorist CBRN-Technical Experts in the Terrorist Group

Y = Total Terrorist CBRN

The model equation appears as: Predicted Amount of CBRN =- 0.4189 + (0.1469) (Political Violence Per Capita Variable) + (1.6197) (Total Number of T-CBRN Technical Experts Variable) + Error

Applying the model

From the multivariate model, we derived an algorithm. The algorithm appears as: ‘Predicted Risk of T-CBRN’ =- 0.4189 + (0.1469) (Political Violence Per Capita Variable) + (1.6197) (Total Number of T-CBRN Technical Experts Variable) + Error

We envision our algorithm being used to answer two questions:

Applying the Algorithm to Question #1: What is the Calculated Risk of T-CBRN occurring in a specific country by nonspecific terrorists?

To determine the answer to this question, specific variables of the country in question would be entered into the algorithm. A calculated risk of CBRN events committed by terrorists in the country would be derived. The calculated risk value derived by the algorithm does not represent the number of CBRN events or casualties predicted to occur by terrorists in the country. The derived risk number represents the amount of risk the country is at for T-CBRN occurring. The amount of risk is interpreted by comparing the country’s calculated risk score with the calculated risk scores of other countries. For example: Country A is at a higher risk of T-CBRN occurring than Country B because Country A has a higher score than Country B.

The two independent variables to be entered into the algorithm are the following: 1. The Degree of Political Violence Per Capita in the Country of Interest (measured by deaths per capita caused by political violence); 2. The Total Number of T-CBRN Technical Experts in the Country of Interest (measured by the total number of known terrorist CBRN-technical experts in the country).

Applying the Algorithm to Question #2: What is the calculated risk of a given terrorist electing to use CBRN weapons?

To determine the answer to this question, specific variables of the lone terrorist/terrorist group in question would be derived from its country of origin and from the terrorist organization. The variable values would then be entered into the algorithm and a calculated risk value would be generated.

Here again, the calculated risk value derived by the algorithm is not to be considered as representing the number of CBRN events or casualties occurring by the given terrorist, but rather to indicate the risk of the terrorist in pursuing CBRN relative to other terrorists which would also be assigned risk scores.

The two independent variables to be entered into the algorithm are the following:

1. The Degree of Political Violence Per Capita in the Terrorist’s Country of Origin, (measured by deaths per capita caused by political violence).

2. The Total Number of T-CBRN Technical Experts in the terroristgroup, (measured by the actual number of known CBRNtechnical experts in the terrorist-group).

Note that the ‘T-CBRN Technical Experts’ variable has changed for addressing this question (question #2), it now entails the total number of CBRN-technical experts in the terrorist-group only. The ‘Political Violence’ variable value is derived from the terrorist’s country of origin. For example: if a terrorist is currently operating in the United States, but its country of origin is Afghanistan, the variable value is derived from Afghanistan since this is assumed to be the main country that contributed to forming the relevant attitudes and philosophies. The ‘political violence’ variable must reflect the culture and conditions that influenced the terrorist’s ideology and formulation of its organization.

Operating the algorithm

To operate the algorithm, the analyst would enter the two variables (depending on which question to be answered) into an algorithm template which would be linked with the statistical coefficients previously derived (below) and the following algorithm would be automatically calculated: Y = α + (b1) (X1) + (b2) (X2)

α = Alpha Coefficient (Y intercept): -0.4189

b1 = Beta Coefficient for Degree of Political Violence Per Capita: 0.1469

b2 = Beta Coefficient for Total Number of T-CBRN Technical Experts: 1.6197

X1 = Total Number of Deaths Caused by Political Violence (in the terrorist’s country of origin/or in the country of interest).

X2 = Total Number of T-CBRN Technical Experts (in the terroristgroup/ or in the country of interest).

Y = Calculated Risk (by the terrorist in question/or in the country of interest).

Testing the algorithm with past T-CBRN events

Example #1: Determining the Calculated Risk of the ‘Taliban’ Pursuing CBRN, Operating in Afghanistan

The Al-Qaeda connected Taliban was originally a traditionalist/ Hanafi Islamist political movement. After being overthrown in 2001 by U.S. forces, it later reassembled in 2004 as a strong insurgency movement. Currently the group is governing mainly the local Pashtun areas of Afghanistan where it continues to wage war (using guerrilla type tactics) against Afghani, Pakistani and North Atlantic Treaty Organization (NATO) led security forces [2].

Using Afghanistan as the Country of Origin, we apply the algorithm:

CR = α + (b1) (X1) + (b2) (X2)

CR = (-0.4189) + (0.1469) (0.3609) + (1.6197) (5)

CR = 7.65, (Confidence band = (±) 0.33)

A calculated Risk of 8 is determined for the Taliban in pursuing CBRN.

Number of CBRN events committed by the Taliban

The Taliban is suspected of being responsible for at least 4 CBRN events: a poison gas attack on a group of female high school students in Kabul (4 May 2010); the threatening to use IEDs laced with Bacillus anthracis against ISAF forces in Tora Bora (1 March 2010); attacking a group of female students in Kandahar with battery acid (12 November 2008), and possessing B. anthracis in the Nangarhar Province (15 January 2007) [2].

Example #2: Determining the Calculated Risk of the ‘Rajneesh’ cult pursuing CBRN, operating in the United States.

In 1984, followers of Bhagwan Shree Rajneesh moved onto the ‘Big Muddy Ranch’ in Wasco County, Dalles, Oregon, where they recruited thousands of U.S. citizens. Soon after the Rajneesh commune was established, the cult took political control of a small nearby town. As a result, relations with the local population quickly dissolved. After being denied building permits, the Rajneesh leadership attempted to gain political control over the rest of the County by influencing a series of County elections [2].

Using the United States as the Country of Origin, we apply the algorithm:

CR = α + (b1) (X1) + (b2) (X2)

CR = 1.12, (Confidence band = (±) 0.33)

A calculated Risk of 1 is determined for the Rajneesh cult in pursuing CBRN.

Number of CBRN events committed by the Rajneesh cult

In 1984, in an attempt to gain political control, Rajneesh cult members launched an attack against 10 local restaurants in Dalles, Oregon with Salmonella enterica Typhimurium resulting in the poisoning of over 750 individuals [2].

Example #3: Determining the Calculated Risk of the ‘Aum Shinrikyo’ cult pursuing CBRN, operating in Japan.

In the early 1990s, the massive Doomsday cult Aum Shinrikyo, led by spiritual leader Shoko Asahar, recruited some of Japans brightest students and scientists. With a billion dollar budget at his disposal, Asahar amassed a number of highly equipped research facilities and equipment conducive for developing CBRN [18].

Using Japan as the Country of Origin we apply the algorithm: CR = α + (b1) (X1) + (b2) (X2)

CR = (-0.4189) + (0.1469) (0) + (1.6197) (800),

CR = 1,295.3, (Confidence band = (±) 0.33)

A calculated Risk of 1,295 is determined for the Aum Shinrikyo cult in pursuing CBRN.

Number of CBRN events committed by the Aum Shinrikyo cult

Aum is considered to have been the most dangerous CBRN-terrorist group ever known [18]. The cult is documented as being involved in at least 45 incidents using CBRN (1989-2000) [2]. Aum is most known for its sarin gas attacks on the Tokyo subway system in 1995 which killed 12 people and injured approximately 6000 [17].

Example #4: Determining the Calculated Risk of the ‘Polisario Front’ pursuing CBRN, operating in Algeria.

In 1973, the nationalist/separatist group the Polisario Front (PF) formed out of discontent from Western Sahara earning its independence from Spain. The group rebelled against the secret dealings that had transpired between the Spanish and Western Sahara pertaining to the governing of the country.

Using Algeria as the Country of Origin we apply the algorithm: CR = α + (b1) (X1) + (b2) (X2)

CR = (-0.4189) + (0.1469) (0.0049) + (1.6197) (0),

CR = -0.50, (Confidence band = (±) 0.33)

A calculated Risk of 0 is determined for the Polisario Front in pursuing CBRN.

Number of CBRN events committed by the Polisario Front

The Polisario Front has only been known to threaten to use CBRNweapons. In 1975, the PF is reported to have instructed the Spanish Basque nationalist group, the ETA ‘to prepare to contaminate the water supplies of Paris, Madrid, Rabat and Nouakchott with Cholera out of retaliation against the policies of France, Spain, and Mauritania toward Western Sahara’ [2].

Example #5: Determining the Calculated Risk of T-CBRN occurring in the ‘United States of America’

The United States of America is the world’s most powerful nation state. For over 50 years, the economy has continued to show steady growth, relatively low unemployment and inflation, and a rapid growth in technological development. With the country’s great growth, there has there also been considerable increase in the number of radical groups and individuals with designs for insurrection through the use of violent tactics. Many of these organizations ascribe to Separatism and Right-Wing ideologies and remain at large today [2,8].

Using the U.S. as the source for the three variables, we apply the algorithm:

CR = α + (b1) (X1) + (b2) (X2)

CR = (-0.4189) + (0.1469) (0.0003) + (1.6197) (12),

CR = 18.94, (Confidence band = (±) 0.33)

A calculated Risk of 19 is determined for T-CBRN occurring in the United States of America.

The Number of T-CBRN events that have occurred in the United States

From compiling documented data on CBRN incidents in the U.S. taken from the James Martin Center for Nonproliferation Studies, approximately 61 CBRN related events are estimated to have occurred thus far by individual perpetrators and/or radical groups [2].

Example #6: Determining the Calculated Risk of T-CBRN occurring in ‘Pakistan’

Pakistan is a federal parliamentary republic consisting of four provinces and four federal territories. Its considerable population spans over 170 million making it the sixth most populated country in the world. With the exception of Indonesia, Pakistan has the largest Muslim population in the world. Because of the country’s partially industrialized economy, it is the 27th largest in the world in respect to purchasing power. Pakistan’s history consists of dark periods of military rule, political turmoil and clashes with its neighboring state India. The country faces a number of challenges which include corruption, poverty, illiteracy and considerable terrorism [8].

Using Pakistan as the source of the three variables we apply the algorithm:

CR = α + (b1) (X1) + (b2) (X2)

CR = (-0.4189) + (0.1469) (0.0002) + (1.6197) (30),

CR = 48.09, (Confidence band = (±) 0.33)

A calculated Risk of 48 is determined for T-CBRN occurring in Pakistan.

Number of T-CBRN events that have occurred in Pakistan

From compiling documented data on CBRN events in the U.S. taken from the James Martin Center for Nonproliferation Studies, approximately 96 CBRN related events are estimated to have occurred by individual perpetrators and/or radical groups [2].

Correlation cluster with T-CBRN

In examining our variables correlating with T-CBRN, we find our best correlations are with National Separatism at (r = 0.8057, p≅ 0.00001, n = 74) (Figure 1), followed by Religious Fundamentalism at (r = 0.7883, p≅ 0.00001, n = 74), CBRN-Technical Experts at (r = 0.7443, p≅ 0.00001, n = 74) (Figure 2), State-CBRN at (r = 0.6891, p≅ 0.00001, n = 74), State-Sponsorship at (r = 0.6489, p≅ 0.00001, n = 74), Political Violence Per Capita at (r = 0.4542, p≅ 0.00001, n = 74) (Figure 3), and finally with Civil Liberties at (r = 0.3107, p < 0.007, n = 74).

Discussion

In evaluating our 12 hypotheses, there were 6 positive findings. Our first finding (hypothesis 2), was that countries with a history of devoting high resources to state-CBRN programs were highly associated with the development of T-CBRN groups (State-CBRN with T-CBRN, (r) = 0.6891, p≅ 0.00001, n = 74). The correlation value suggests that there is a connection with CBRN-terrorist groups developing in countries that currently have or have had at one time state funded CBRN programs. We believe this connection is due to a number of factors such as in the availability of trained technical personnel in the CBRN-technologies, as well as the availability to seed stocks, chemicals, laboratories, equipment etc. in the country. In addition to the factor of the availability of technical resources from CBRN-countries, intrinsic ideological factors could also be involved. These ideological factors may be due to the citizens having a culturally inherent appreciation of the feasibility of CBRN, drawn from the knowledge of past state successes and propaganda.

Our second positive finding (hypothesis 4), was that countries having high political violence per capita were associated with T-CBRN activity (Political Violence Per Capita with T-CBRN, (r) = 0.4542, p < 0.0004, n = 74) (Figure 3). The correlation value suggests that there is a connection with political violence in a country and T-CBRN activity that develops within it. We believe the connection may be due to the influence a highly violent culture has on the attitudes of its citizens and terrorists that come from these cultures. As was the case with CBRN countries, here we also detect an intrinsic appreciation which in this case, is for the use for violent tactics. Although the appreciation by citizens may only be at a tolerance level for political violence, terrorists in these cultures seem to embrace the tactic. We believe that for terrorists coming from highly violent cultures in which violent acts are a way of life, it would be difficult for them to adopt different, less violent tactics, since violence is their way of life. The violent cultures terrorists come from may serve as a model on which to base their tactics. Post states that terrorists most likely to seek CBRN are those that wish to inflict high numbers of casualties [16]. As we know, the use of CBRN would be a way of achieving this objective. We believe our negative finding from our related political violence hypothesis (hypothesis number 3), which measured the raw score of political violence in a country, failed due to its being an entirely different variable. By measuring ‘per capita’ (deaths caused by political violence/population) we were able to put the degree of political violence in a country in proper perspective which revealed a considerably harsher environment.

Our third positive finding (hypothesis 5), suggests that countries with terrorist groups that have members that are CBRN-technical experts are highly connected with T-CBRN activity occurring within it (CBRN-Technical Experts with T-CBRN, (r) = 0.7443, p≅ 0.00001, n = 74) (Figure 2). Our correlation value showed a considerably strong connection between T-CBRN and T-CBRN technical experts. This suggested that the amount of CBRN-technical experts in a terrorist group is an important indicator as to whether the group may choose to pursue CBRN. We believe this variable to be an important indicator of a terrorist’s propensity for seeking CBRN. The Millennial-cult Aum Shinrikyo had approximately 800 technical experts developing CBRN, and used their weapons repeatedly [17]. The Rajeneesh had 4 biological experts, and it too used CBRN repeatedly poisoning over 700 people. We speculate that technical personnel, due to their knowledge, would have considerable influence on a terrorist group’s decision making, including the choice of weaponry. Not only would CBRN experts likely be in leadership positions, but they would also be equipped with the needed skills, knowledge and connections to develop CBRN. Furthermore we believe their backgrounds would bias them to develop CBRN weapons since it would be an area they would be very familiar with.

Our fourth positive finding (hypotheses 6), suggests that countries with CBRN-terrorists aspiring to ‘Religious Fundamentalism’ ideologies are highly associated with T-CBRN (Religious Fundamentalism Ideology with T-CBRN, (r) = 0.7883, p≅ 0.00001, n = 74). Post states that the groups most likely to use CBRN in their tactics are those that aspire to religious ideologies [16]. From the correlation value it was evident that Religious Fundamentalist groups are associated with CBRN. Religious Fundamentalist groups have proven to be extremely dangerous in the past and are known to have acquired and used CBRN. In hypothesis number five we discussed how technical-knowhow is a major event in T-CBRN. In hypotheses number 6, we have now confirmed a strong ideological connection with T-CBRN with our strong correlation value. In 2007, it was seen just how dangerous the combination of religiousfundamentalism and CBRN technical expertise can be, when an Al- Qaeda connected group detonated multiple chlorine tanker trucks in Ramadi and Fallujah, Iraq [2].

Our fifth positive finding (hypothesis number 10), suggests that foreign countries with a high number of U.S. troops present are associated with T-CBRN (Number of U.S. Troops with T-CBRN, (r) = 0.2340, p < 0.04, n = 74). From the correlation value of hypothesis number 11 ‘Troops Per Capita’, we found no strong association. We believe that this is due to the fact that it is the actual raw total that is significant, not the per capita measure. From our correlation result, we surmise that occupation of countries by U.S. troops is associated with terrorist resentment and use of CBRN. This was not a surprising finding. Over the past 20 years, we have seen the incidence of terrorism rise with the escalation of U.S. troops in Lebanon, Saudi Arabia, Iraq, Afghanistan, Pakistan, etc. [2,19-22,25].

Our sixth positive finding (hypothesis 12) suggests that countries that sponsor terrorist groups are highly associated with T-CBRN occurring within them (State-Sponsor Potential with T-CBRN, (r) = 0.6489, p≅ 0.00001, n = 74). From the high correlation value, we see that there is a strong association with states that have both CBRN capability and a history of sponsoring terrorism and T-CBRN. We explain this correlation by proposing two scenarios, either corrupt government officials are supplying terrorists with CBRN capability for their personal gain, or it is the state itself that’s equipping terrorists with CBRN technology. Theoretically, the latter scenario would entail a state government supplying terrorists with some type of CBRN for carrying-out specific objectives. These objectives would likely include the targeting of specific groups of citizens within its own country or in foreign states. Terrorist groups would essentially be contracted by a state to perform these types of acts in order to create ‘plausible deniability’ [22].

A negative finding occurred with electricity production and T-CBRN (hypothesis 1). Electricity production did not seem to be a major element for T-CBRN. It may be that some of the countries having low electricity production still yielded enough power to operate any special equipment that would have been required for CBRN production.

Our second negative finding was with countries with CBRNterrorist groups aspiring to Right-Wing ideologies. Jerald Post reports that groups aspiring to Right-Wing philosophies were likely to pursue CBRN should they be able to obtain them. We however, did not find an association to exist here. Perhaps CBRN capability is still beyond many of these groups [16]. Our conclusion drawn from biographical data regarding the memberships of many of these groups is that much of the membership does in fact lack expertise in the CBRN fields. This may explain why CBRN activity is low overall with these types of groups.

Negative findings were also observed with our two hypotheses 8 and 9, speculating on a country’s propensity in producing terrorists that would pursue CBRN. Asal and Rethemeyer suggest that cultural integration into the westernized world could potentially give terrorists an enhanced awareness of what might be achievable, but concede that their findings did not bear this out. Their findings actually indicated that democratic regimes were less likely to host organizations that seek CBRN capabilities [7]. In testing their variable, we actually found no association to exist at all between our western embeddedness variables (the raw Number of McDonald’s Restaurants, and Number of McDonald’s Restaurants Per Capita). Perhaps the number of McDonald’s Restaurants is not an accurate measure of true western influence in a country. It’s been suggested that using this model to represent religious and political philosophies would be flawed due to it actually being a business model consisting of business oriented, capitalistic strategies and interests.

Conclusion

Correlative analysis can potentially provide a precise approach for predicting terrorist CBRN. In this study, however, there were limitations in the quality of T-CBRN data available. This deficiency existed primarily with some of the CBRN related events such as in attempts to construct or deploy CBRN weapons being reported to have been committed by unknown terrorists. Because of the inability of authorities to identify perpetrators responsible for some of the CBRN activity recorded on the databases we drew our variables from, the incidents had to be excluded from our dataset. Other areas in which we were limited were in the available biographical data of terrorists, such as in their education and technical experience. Financial data were also limited for terrorists. Nevertheless, despite these limitations, we have shown our Random Nations approach to be statistically accurate in its description of T-CBRN activity.

From our bivariate analyses, we found strong correlations to exist among variables associated directly and indirectly with T-CBRN activity. For example, countries with state-CBRN programs and a large number of T-CBRN technical experts with a history of political violence are especially vulnerable to T-CBRN activity.

The next step

As intelligence capabilities improve, so will the predictive power of our model and future models that will one day be developed. Future studies should focus on expanding the Random Nations Matrix to include all of the countries of the world. As new data become available, researchers will be able to investigate more comprehensively the motivations of terrorists in choosing CBRN. For example, complete biographical data of group members would have greatly enhanced our technical variable, ‘Number of CBRN-Technical Experts’. Another area in which more precise data would help considerably would be in the financial resources of a terrorist group. We believe that an important factor of T-CBRN is the amount of money a group or lone terrorist would have at its disposal. This was clearly seen as a factor with Aum Shinrikyo in its immense spending in developing various types of CBRN weapons. Unfortunately, the data available in this area is grossly incomplete. As intelligence gathering continues to improve, new variables may become available. It is likely that in addition to these new variables, increased accuracy in some of the existing variables already examined would also result, such as with those concerning technical expertise, funding and regional ideologies; in our opinion, very important factors which should be further explored.

References

1. Center for Nonproliferation Studies (2009)
2. Center for Nonproliferation Studies (2011)
3. Guthrie R (2010)
4. National Infrastructure Protection Center (2003)
5. Parachini J (2003)
6. Sullivan TJ, WL Perry (2004)
7. Asal VH, Rethemeyer RK (2009)
8. Central Intelligence Agency (2009)
9. Center for Defense Information (2009)
10. Freedom House (2009)
11. Center for Systemic Peace (2009)
12. World wide military deployments (2010)
13. McDonald's Corporation (2010) List of countries with McDonald's franchises. McDonald's Corporation,Oak Brook, IL, USA.
14. Nissan, R.-K. (2009)
15. Alibek K, S Handelman (1999)
16. Post JM (2002)
17. Kaplan DE (1996)
18. Public Broadcasting Service (2010)
19. Bergen P (2006)
20. National Memorial Institute for the Prevention of Terrorism (2006)
21. Frontline (2001)
22. Quillen C (2000)
23. Knoke D, G Bohrnstedt (2002) Statistics for social data analysis. Wadsworth Publishing.
24. Kachigan S (1991)
25. Jalali AA (2001)

--