1. Introduction

Every year a lot of aviation incidents and accidents happen all around the World in which different airlines are involved, with and without fatalities and including different type of airplane. As highlighted in (Oster, Strong & Zorn, 2013), traditionally the focus of research on aviation safety has been on analyzing accidents and correlated factors, investigating possible root causes, and recommending specific corrective measures and action. Furthermore, statistical methods are proven very useful in finding the patterns and connections between the various data available for every individual event (either incident or accident).

Aim of this report is to analyse the incidents and accidents from the last 2017 and determine if the statistically significant connection exists among the data. For the purpose of this analysis dataset is obtained from the aviation safety website (ASN, 2017) and given in the appendix 1. Main goal is to determine if the number of fatalities are correlated with the aircraft type and geographical location. Second goal is to create a regression model for predicting the number of fatalities based on the all available variables in the dataset.

Decision variables:
Variable of interest: Number of fatalities [NOF]

Influence variables:
Type of Aircraft: Airbus [TAA], ATR [TAT], Boeing [TAB], Cessna [TAC], Other [TAO].
Geographical location: Africa [AF], Asia [AS], Europe [EU], N. America [NA], S. America [SA], Oceania [OC], Ocean [OO].

Other available variables to be used for prediction model:
Accident category: Hull-loss [ACH], Repairable damage [ACR].
Accident type: Accident [A], Incident [I], Other occurrence [O].
Season: Winter [W], Summer [S].

For the purpose of this analysis next hypotheses are formulated:

For goal 1: determine if the number of fatalities are correlated with the type of the aircraft
H0: μTAA=μTAT=μTAB=μTAC=μTAO (there is no effect of aircraft type on number of fatalities)
H1: Fatalities means for different aircraft types are not all equal; at least one differs from the others.

For goal 2: determine if the number of fatalities are correlated with the geographical location
H0: μAF=μAS=μEU=μNA=μSA=μOC=μOO (there is no effect of geog. location on number of fatalities)
H1: Fatalities means for different geog. locations are not all equal; at least one differs from the others.

2. Methods

A section on Methods where the data collection method and the analytical technique are described. This section should consist of about a paragraph in which you describe your method.

First, descriptive statistics will be calculated for the variable of interest - Number of fatalities, using the influence variables - accident category, accident type and season. After that, for each of the goals next will be conducted, step by step:

Step 1. Single factor analysis of variance (ANOVA) for the variables of interest using the influence (explanatory) variables - Type of Aircraft and Geographical location.
Step 2. Tukey Test for further comparison of the differences in means (if H0 is rejected).
Step 3. Regression model for predicting the number of fatalities based on the variables.

3. Results

Complete descriptive statistics for fatalities is provided in the appendix 2.

We are interested in the between group source of variation. For the goal 1, as presented in table 2. F statistic show that there is a large amount of variance in fatalities between the type of aircraft. Since F > Fcrit, (2.86>2.38) we can reject the null hypothesis. The fatalities means for the different type of aircraft are not equal and at least one differs from the others. However, since the p-value (0.02) is < 0.05, further test is required. For the goal 2, as presented in table 3. F statistic is indicating that there is no significant variance present in fatalities between the geographical locations. Since F < Fcrit, (1.4<2.13) we can accept the null hypothesis. The fatalities means for the different geographical locations are equal. However, since the p-value (0.21) is > 0.05, further test is required. Complete results from the ANOVA

1. While an integrated ERM environment is a much sought-after goal,...