通过具有相同含义的经济指标序列对同一个问题进行研究时，使用不同频率的指标所得到的结果可能会存在一定差异。因此，在实际研究中，需根据研究目的的差异性选取不同频率的指标序列，并基于已有数据，通过较为科学的频率转换方法，实现各指标序列间频率的统一性。一般来说，指标序列的频率转换方法包含由高频向低频转换和由低频向高频转换两个方向。其中，对于低频的经济指标序列而言，将其转换为对应的高频序列，能够得到包含更多信息的分析结果，并为建模方法的选择提供更多可能性。

首先，本文对常见频率转换方法的基础理论及其发展进行了系统性梳理。包含阶梯函数方法、线性插值方法等传统方法，以及Denton方法、Litterman方法、Chow-Lin方法等现阶段较为主流的频率转换方法。接下来，本文基于相关理论与实践成果，创建了R包“lowtohigh”，通过“DCLFC”和“batchDCLFC”两个函数实现了一个或一组序列由低频到高频的转换，并有效解决了EViews软件在频率转换过程中存在的部分问题。同时，本文就如何通过EViews和R两种软件实现序列由低频向高频的频率转换过程进行了较为详细的说明，并以MRE和相关系数两个指标作为评判标准，验证了通过两种软件能够得到较为一致的结果。基于此，本文选取了我国及部分省市地区的名义GDP、城镇登记失业人员数等包含流量、存量、指数三种数据类型，年度、季度、月度三种数据频率的多个指标序列作为数据集。在对低频目标序列平稳性、低频目标序列长度、转换频率需求、地区等因素分别控制的情形下，得出了非平稳的低频目标序列应当选取较为合适的高频相关序列作为参照、低频序列长度对频率转换效果的影响应当根据序列是否平稳分情况讨论、可以基于已知数据对合适的频率转换方法和相关高频序列进行探索等结论。最后，本文从 R代码的优化和频率转换方法的优化两个层次，对今后更进一步的研究方向。

Generally, when the same problem is researched through the same meaning of economic index series, the results obtained by using different frequency of indicators may be different. Therefore, in practical research, it is necessary to select indicator series with different frequencies according to the difference of research purposes, and to ensure the uniformity of frequencies among the indicator series through a frequency conversion method based on the existing data. Generally speaking, the frequency conversion method of indicator series contains two directions: conversion from high to low  and conversion from low  to high. Among them, for the low-frequency economic indicator series, converting them to the corresponding high-frequency series can get the analysis results with more information, and provide more possibilities for the choice of modeling methods.

At first, We include traditional methods such as step function method, linear interpolation method,  and more frequency conversion methods including Denton method, Chow-Lin method and Litterman method. Next, based on the relevant theoretical and practical results, we create the R package "lowtohigh", and implements the conversion of a series or a group of series from low to high frequency by the function "DCLFC" and "batchDCLFC", it effectively solve some  problems that exist in EViews software in the process of frequency conversion. At the same time, we provides a detailed description of the process of frequency conversion from low to high frequencies by EViews and R software. Then, we use MRE and correlation coefficient as the evaluation criteria to judge the effect of frequency conversion, and verifies that we can get consistent results by the two software. Based on that, we select a number of indicator series containing three data types of flow, stock and index, and three data frequencies of annual, quarterly and monthly, such as nominal GDP, the number of urban registered unemployed persons of China and some provinces as data sets. In the case of controlling for the smoothness of the low-frequency series, the length of the low-frequency series, the conversion frequency demand, and the region separately, it is concluded that the non-smooth low-frequency series should select a more suitable high-frequency correlation series as a reference, and the effect of the low-frequency series length on the frequency conversion effect should be discussed in separate cases according to whether the series is smooth or not, and the suitable frequency conversion method and the relevant high-frequency series can be explored based on known data. Finally, we provide an outlook on further research directions from the optimization of R codes and the optimization of frequency conversion methods.