In this paper, we introduce a matrix quantile factor model for matrix sequence data analysis. For matrix-valued data with a low-rank structure, we estimate the row and column factor spaces via minimizing the empirical check loss function over all panels. We show that the estimates converge at rate $1/\min\{\sqrt{p_1p_2}, \sqrt{p_2T}, \sqrt{p_1T}\}$ in the sense of average Frobenius norm, where $p_1$, $p_2$ and $T$ are the row dimensionality, column dimensionality and length of the matrix sequence, respectively. This rate is faster than that of the quantile estimates via ``flattening" the matrix quantile factor model into a large vector quantile factor model, if the interactive low-rank structure is the underground truth. We provide three criteria to determine the pair of row and column factor numbers, which are proved to be consistent. Extensive simulation studies and an empirical study justify our theory.
翻译:在本文中, 我们引入矩阵序列数据分析的矩阵量化系数模型。 对于结构低的矩阵值数据, 我们通过将所有面板的经验检查损失函数最小化来估计行和列系数空间。 我们显示, 如果交互式低位结构是地下真理, 则该估计数以1美元/ min ⁇ sqrt{ p_ 2T}, \ sqrt{ p_ 1T ⁇ $为平均法兰比纽斯标准值, 其中美元/ 1美元、 美元/ 2美元和美元/ 美元为矩阵序列的行维度、 列维度和长度。 这个比率比通过“ 缩放” 矩阵微分系数模型算成一个大的矢量系数模型的四分位估计速度要快。 我们提供三个标准来确定行数和列因数的对数, 并证明它们是一致的。 广泛的模拟研究和一项经验研究证明了我们理论的合理性。