以電訪樣本資料庫作為解決原住民調查研究難題之可行性的再討論

Using Telephone Survey Sample Data asa Feasible Resolution to the Difficulty of Indigenous Survey

國內以原住民為對象之調查研究較為缺乏，且普遍為附屬於全國抽樣當中的小樣本分析，其主因在於台灣原住民人口數量較少，約僅占全國總人口數的2％，輔以其地理分布相當不均勻，以及相當數量的原住民因為生計因素而有工作／居住地與戶籍所在地不同的情形，致使傳統調查方式，無論面訪或電訪，都有難以解決的抽樣困難。然而近年來，國內民意調查日益普遍，學術與民間調查機構已經累積數量極為龐大的電話調查成功樣本，其中包含大量的漢人樣本與一定比例的原住民樣本。若能取得這些「原本不是以原住民為母體，卻包含原住民的成功樣本」，將混雜埋藏於大量「無用」資料當中的「有價值資料」給挖掘出來，應能直接發掘篩選出所需要的原住民樣本，節約成本而解決前述抽樣困難。有鑑於此，研究團隊與全國公信力民意調查股份有限公司合作，合併2007年迄今（2013年3月）該公司所執行抽樣範圍為全國的原住民電話訪問成功樣本成為原住民電話資料庫，以為調查之用。初步研究結果顯示，使用雙重抽樣方法所整併的電訪樣本資料庫進行電話調查，所獲得之成功樣本特徵與官方人口統計中的原住民母體特徵近似。同時，納入跨年度的原住民電話資料，能夠更有效地提升前述原住民電話資料庫的蒐集範圍，並降低樣本取得成本。

There is a lack of indigenous object research inside the nation, and the research is usually attached to national sample as one of the small sample analysis. The main reason is that with only about 2% of the country's total population, indigenous population is rather few in Taiwan. Besides, uneven geographic distribution and the factor that a considerable number of workplace/residence and domicile difference due to livelihood are other reasons which cause difficulty to traditional method of investigation. Both face-to-face interview or telephone interview have handicap on sampling. However, public opinion polls had become more universal in recent years. Academic and private investigation agency all has accumulated a large number of successful phone survey samples, which contained numerous Han samples and certain proportion of indigenous samples. If there is a possibility to get these samples which were 'not originally using indigenous population as statistic population, but contain successful indigenous samples,' and apply data mining technology to dig the 'valuable mineral' out from a large amount of 'useless' information, then it will be probable to directly filter out the indigenous samples that were needed, and solve the problems mentioned above in the context of saving cost. In this view, the research team and Taiwan Real Survey Co. had cooperated and merged the successful indigenous population phone interview samples, which are from national samplings that were executed by the company since 2007 so far (March 2013), as indigenous telephone database. Preliminary research result shows, by using telephone interview sample database that was consolidated by data mining and double sampling, the characteristics of successful samples which got from the telephone interview are similar to the indigenous statistic population in the official population statistics. Other than that, including multi-year indigenous population phone data could enhance the collection range of the indigenous telephone database, and reduces the cost of sample acquisition.