A group divisible design GD (k, λ, t; tu ) isα-resolvable if its blocks can be partitioned into classes such that each point of the design occurs in precisely α blocks in each class. The necessary conditions for the existence of such a design are λt( u – 1) = r ( k – 1) , bk=rtu; k│αtu andα│ r. It is shown in this paper that these conditions are also sufficient when k =3, with some definite exceptions.

We shall refer to a strong partially balanced design SPBD(v, b, k; λ, 0) whose b is the maximum number of blocks in all SPBD(v, b, k; λ, 0), as an optimal strong partially balanced design, briefly OSPBD(v, k, λ). The author inpaper (Discrete Math. 279 (2004) 173) investigated the existence of OSPBD(v, 5, 1) and gave the spectra of OSPBD(v, 5, 1) for v ≡ 0, 1, 3 (mod 4). Inthis article we shall investigate the existence of OSPBD(v, 5, 1) and give the spectrum of OSPBD(v, 5, 1) for the remaining case v ≡ 2 (mod 4).

Let Km,n be a complete bipartite graph with two partite sets having m and n vertices, respectively. A Pv-factorization of Km,n is a set of edge-disjoint Pv-factors of Km,n which partition the set of edges of Km,n. When v is an even number, Wang and Ushio gave a necessary and sufficient condition for existence of Pv-factorization of Km,n. When k is an odd number, Ushio in 1993 proposed a conjecture. Very recently, we have proved that Ushio’s conjecture is true when v = 4k − 1. In this paper we shall show that Ushio Conjecture is true when v = 4k +1, and then Ushio’s conjecture is true. That is, we will prove that a necessary and sufficient condition for the existence of a P4k+1-factorization of Km,n is (i) 2km ≤ (2k + 1)n,(ii) 2kn ≤ (2k +1)m, (iii) m+n ≡ 0 (mod 4k +1), (iv) (4k +1)mn/[4k(m+n)] is an integer.

In this paper it is shown that an idempotent TD (5, m) − TD (5, n) exists whenever the known necessary condition m ≥ 4n+1 is satis3ed, except when (m, n)=(6, 1) and possible when (m, n)=(10, 1). For m ＜ 60 and n ≤ 10, we also indicate where several idempotent TD(k, m)−TD(k, n)’s for k = 6, 7 can be found.

We prove that there exists a pair of orthogonal diagonal Latin squares of order v with missing subsquares of side n (ODLS (v, n ) for all v≥3n+2 and v-n even. Further, there exists a magic square of order v with missing subsquare of side n (MS (v, n)) for all v≥3n+2 and v-neven.