The document discusses arithmetic triangles and ARMA models.
1. The arithmetic triangle, also known as Pascal's triangle, is a triangular array where each number is the sum of the two numbers directly above it. It was studied by mathematicians like Pascal, Tartaglia, and al-Kashi.
2. ARMA models are used in time series analysis and consist of autoregressive (AR) and moving average (MA) parts. The AR part models a time series as a linear combination of past observations plus noise. The MA part models a time series as a linear combination of past noise values.
3. ARMA models combine both AR and MA parts, representing processes generated from past
The document discusses arithmetic triangles and ARMA models.
1. The arithmetic triangle, also known as Pascal's triangle, is a triangular array where each number is the sum of the two numbers directly above it. It was studied by mathematicians like Pascal, Tartaglia, and al-Kashi.
2. ARMA models are used in time series analysis and consist of autoregressive (AR) and moving average (MA) parts. The AR part models a time series as a linear combination of past observations plus noise. The MA part models a time series as a linear combination of past noise values.
3. ARMA models combine both AR and MA parts, representing processes generated from past
The document discusses arithmetic triangles and ARMA models.
1. The arithmetic triangle, also known as Pascal's triangle, is a triangular array where each number is the sum of the two numbers directly above it. It was studied by mathematicians like Pascal, Tartaglia, and al-Kashi.
2. ARMA models are used in time series analysis and consist of autoregressive (AR) and moving average (MA) parts. The AR part models a time series as a linear combination of past observations plus noise. The MA part models a time series as a linear combination of past noise values.
3. ARMA models combine both AR and MA parts, representing processes generated from past
200 BC. A Between the fourteenth and the fifteenth cen- turies, al-Kashi, a mathematician from the Iranian city of Kashan, wrote The Key to Arithmetic. In this work he calls binomial For example: coefficients “exponent elements”. C64 = C53 + C54 = 10 + 5 = 15 . In his work Traité du Triangle Arithmétique, published in 1665, Pascal, Blaise (1654) defined the numbers in the “arithmetic tri- angle”, and so this triangle is also known as Pascal’s triangle. We should also note that the triangle was made popular by Tartaglia, Niccolo Fontana in 1556, and so Italians often refer to it as Tartaglia’s triangle, even though Tartaglia More generally, we have the relation: did not actually study the arithmetic triangle. k+1 Cnk + Cnk+1 = Cn+1 , MATHEMATICAL ASPECTS because: The arithmetic triangle has the following form: n! Cnk + Cnk+1 = (n − k)! · k! 1 n! 1 1 + (n − k − 1)! · (k + 1)! 1 2 1 n! · [(k + 1) + (n − k)] = 1 3 3 1 (n − k)! · (k + 1)! 1 4 6 4 1 (n + 1)! = 1 5 10 10 5 1 (n − k)! · (k + 1)! k+1 1 6 15 20 15 6 1 = Cn+1 . ... FURTHER READING Each element is a binomial coefficient Binomial n! Binomial distribution Cnk = Combination k! (n − k)! n · (n − 1) · . . . · (n − k + 1) Combinatory analysis = . 1· 2·...·k This coefficient corresponds to the element REFERENCES k of the line n + 1, k = 0, . . . , n. Pascal, B.: Traité du triangle arithmétique Any particular number is obtained by adding (publ. posthum. in 1665), Paris (1654) together its neighboring numbers in the pre- Pascal, B.: Œuvres, vols. 1–14. Brun- vious line. schvicg, L., Boutroux, P., Gazier, F. (eds.) 20 ARMA Models
Paris (1904–1925) + θp yt−p + εt , Pascal, B.: Mesnard, J. (ed.) Œuvres com- where θ1 , θ2 , . . . , θp are the positive or plètes. Vol. 2. Desclée de Brouwer, Paris negative parameters to be estimated and (1970) εt is the error factor, which follows a nor- Rashed, R.: La naissance de l’algèbre. In: mal distribution. Noël, E. (ed.) Le Matin des Mathémati- 2. MA model (moving average) ciens. Belin-Radio France, Paris (1985) In a moving average process of order q, Chap. 12). each observation yt is randomly generat- Youschkevitch, A.P.: Les mathématiques ed by a weighted arithmetic mean until arabes (VIIIème-XVème siècles). Partial the qth period: translation by Cazenave, M., Jaouiche, K. MA(1) : yt = εt − α1 εt−1 Vrin, Paris (1976) MA(2) : yt = εt − α1 εt−1 − α2 εt−2 ···
ARMA Models MA(p) : yt = εt − α1 εt−1 − α2 εt−2
− . . . − αq εt−q , ARMA models (sometimes called Box- Jenkins models) are autoregressive moving where α1 , α2 , . . . , αq are positive or nega- average models used in time series analy- tive parameters and εt is the Gaussian ran- sis. The autoregressive part, denoted AR, dom error. consists of a finite linear combination of The MA model represents a time series previous observations. The moving aver- fluctuating about its mean in a random age part, MA, consists of a finite linear manner, which gives rise to the term combination in t of the previous values for “moving average”, because it smoothes a white noise (a sequence of mutually inde- the series, subtracting the white noise gen- pendent and identically distributed random erated by the randomness of the element. variables). 3. ARMA model (autoregressive moving average model) ARMA models represent processes gen- MATHEMATICAL ASPECTS erated from a combination of past values 1. AR model (autoregressive) and past errors. They are defined by the In an autoregressive process of order p, following equation: the present observation yt is generated by a weighted mean of the past observations ARMA(p, q) : up to the pth period. This takes the follow- yt = θ1 yt−1 + θ2 yt−2 + . . . ing form: + θp yt−p + εt − α1 εt−1 − α2 εt−2 AR(1) : yt = θ1 yt−1 + εt , − . . . − αq εt−q , AR(2) : yt = θ1 yt−1 + θ2 yt−2 + εt , with θp = 0, αq = 0, and (εt , t ∈ Z) is .. . a weak white noise.