大考中心的試題在正式考試之前並未施測，試題難易度是依據編製試題者的主觀預估而定，若能使其熟悉試題特徵對考生作答的影響，便能有效控制試題難易度。以往的研究指出試題特徵為解題過程中所牽涉的各個變項，各變項可當成難易度預估的因素指標。數學科的研究指出難易度指標為學科內容、思考策略、步驟數與熟悉度，但未量化各因素指標對答對率的影響。本研究採用數學科所建議的試題結構分析方法，對物理試題的物理概念內容、物理概念數目、數學技巧、思考策略與測驗層次等五個項目，進行試題結構分析。本研究分析的34道試題幾乎皆測驗力學中「動力學」及「功與能量」兩單元相關概念，著重於分析相同概念下其他因子對試題難易度的影響，不討論學科內容對難易度的影響。依據試題結構分析結果，本研究使用概念數、數學技巧、創新性與測驗層次等四個因素指標，並用兩種模式計算預估的答對率：一則是由各指標值加總後的平均來推算，另一則是對各因素指標乘上不同權重後的總和來推算。結果顯示各因素指標加以不同權重，能使預估答對率較接近實測結果，這樣的結果顯示各因素指標對難易度的影響可能不是等值的。雖然答對率預估值的精確度還有改進空間，本研究提出了一種難易度各因子量化的方法，以及用各因素指標計算答對率的模型，期能為往後物理試題難易度預估提供研究的基礎。

Tests in the College Entrance Examination Center are not pre-tested before administration. As a result, difficulty of the tests is subjectively estimated by test writers. If we could learn the impact of test item characteristics on the candidates, it will be easier to control test difficulty. Previous studies have pointed out that test characteristics involve variables of problem solving. Studies on Math tests have found four major factors that exert influence on test difficulty: subject content, thinking tactics, number of solution steps, and familiarity with the problem. All these factors can be considered predictors for difficulty. However, these studies did not quantify the factors so as to figure out the probability of getting correct answers. In this study, we adopted the test question structure analysis method, as used in Math tests studies, and analyzed test questions on the basis of physics concepts, the number of physics concepts, mathematical skills, thinking tactics and test levels. The study analyzed 34 questions in the Physics test that address the content of "Dynamics" and "Work and Kinetic Energy" in the field of Mechanics. The focus was on how the abovementioned factors influence the difficulty of questions that contain the same physics concepts. According to the results of the question structure analysis, the number of physics concepts, math skills, innovation and test levels were used as predicting factors. Two models were employed to calculate the estimated percentage of correctness. One model calculated the percentage by the average of all factor indexes; the other obtained the percentage by giving different factors different weights. The factors with different weights were found making better prediction. It could be inferred that all factors might be of different magnitudes. Although there are still rooms to improve the accuracy of the estimated percentage, the current study has proposed a method of quantification of various factors. It is hoped that it could be used in further studies on predicting difficulty of the Physics test.