泊松分佈(Poisson distribution)是一經常被配適於稀有事件建模的機率分配，其應用領域相當的廣泛，如生物，商業，品質控制等。其中許多的應用均為兩群體均數的比較，如欲檢測一新的處理是否較原本的處理俱優越性(superiority)，或者欲驗證一新的方法相較於舊的方法是否俱有不劣性(non-inferiority)。因此，此研究的目標為發展假設檢定的方法，用於比較兩獨立的泊松樣本是否有優越性及不劣性。一般探討假設檢定方法時，均因干擾參數的出現而導致理論探討及計算上的困難。為因應此困境，本研究由簡入繁，亦即先探討相等式的虛無假設(the null hypothesis of equality)，繼而，再推展至非優越性的虛無假設(the null hypothesis of non-superiority)，最後將這些探究的假設檢定方法應用至檢定不劣性並驗證這些方法的適用性。 兩種Wald 檢定統計量是本研究主要的研究興趣。對應於這兩種檢定統計量的近似的假設檢定法，是利用其極限分配為常態分配的特性而衍生的。此研究裡，可推導得到近似的檢定法的檢定力函數及欲達成某一檢定力水平時所需的樣本數公式。並依據此檢定力函數檢驗此檢定法的有效性(validity)及不偏性(unbiasedness)。並且推廣一連續修正的方法至任何的樣本數組合。另外一方面，此研究亦介紹並推廣兩種p-值的正確(exact)檢定法。其中一種為信賴區間p值檢定法(Berger和Boss, 1994), 另一種為估計的p值檢定法(Krishnamoorthy和Thomson, 2004)。一般正確檢定法較需要繁瑣的計算，故此研究將提出某些步驟以降低計算的負擔。就信賴區間p值檢定法而言，其首要工作為縮減求算p值的範圍，並驗證所使用的檢定統計量是否滿足Barnard凸面(convexity)的條件。若此統計量符合凸面convexity的條件，且在Poisson 的問題上，則此正確的信賴區間p值將出現在屬於虛無假設的參數空間的邊界上。然而，對於估計的p值檢定法而言，因在虛無假設的參數空間上求得Poisson均數的最大概似估計值，並不簡單及無法直接求得，故在此研就，將以一Poisson均數的點估計值代替。對於正確的假設檢定方法，此研究亦提出一個欲達成某一檢定力水平時所需的樣本數的步驟。 此研究將透過一個廣大的數值分析來驗證之前所提出的假設檢定方法。其中，可發現這些近似的假設檢定法之間的差異會受到兩群體之樣本數之比率的影響，而連續性的修正於某些情況下確實能夠使型I誤差較能夠受到控制。另外，當樣本數不夠多時，正確的假設檢定法是較近似的假設檢定法適當，尤其在型I誤差的控制上更是明顯。最後，此研究所提出的假設檢定方法將實際應用於一組乳癌治療的資料。 / The Poisson distribution is a well-known suitable model for modeling a rare events in variety fields such as biology, commerce, quality control, and so on. Many applications involve comparisons of two treatment groups and focus on showing the superiority of the new treatment to the conventional one, or the non-inferiority of the experimental implement to the standard implement upon the cost consideration. We aim to develop statistical tests for testing the superiority and non-inferiority by two independent random samples from Poisson distributions. In developing these tests, both computational and theoretical difficulties arise from presence of nuisance parameters. In this study, we first consider the problems with the null hypothesis of equality for simplicity. The problems are extended to have a regular null hypothesis of non-superiority next. Subsequently, the proposed methods are further investigated in establishing the non-inferiority. Two types of Wald test statistics are of our main research interest. The correspondent asymptotic testing procedures are developed by using the normal limiting distribution. In our study, the asymptotic distribution of the test statistics are derived. The asymptotic power functions and the sample size formula are further obtained. Given the power functions, we justify the validity and unbiasedness of the tests. The adequate continuity correction term for these tests is also found to reduce inflation of the type I error rate. On the other hand, the exact testing procedures based on two exact $p$-values, the confidence-interval $p$-value (Berger and Boos (1994)), and the estimated $p$-value (Krishnamoorthy and Thomson (2004)), are also applied in our study. It is known that an exact testing procedure tends to involve complex computations. In this thesis, several strategies are proposed to lessen the computational burden. For the confidence-interval $p$-value, a truncated confidence set is used to narrow the area for finding the $p$-value. Further, the test statistic is verify whether they fulfill the property of convexity. It is shown that under the convexity the exact $p$-value occurs somewhere of the boundary of the null parameter space. On the other hand, for the estimated $p$-value, a simpler point estimate is applied instead of the use of the restricted maximum likelihood estimators, which are less straightforward in this problem. The estimated $p$-value is shown to provide a conservative conclusion. The calculations of the sample sizes required by using the two exact tests are discussed. Intensive numerical studies show that the performances of the asymptotic tests depend on the fraction of the two sample sizes and the continuity correction can be useful in some cases to reduce the inflation of the type I error rate. However, with small samples, the two exact tests are more adequate in the sense of having a well-controlled type I error rate. A data set of breast cancer patients is analyzed by the proposed methods for illustration.

User retention is a key factor for Software as a Service (SaaS) companies to ensure long-term growth and profitability. One area which can have a lasting impact on a digital product’s user retention is its user onboarding experience, that is, the methods and elements that guide new users to become familiar with the product and activate them to become fully registered users. Within the area of user onboarding, multiple authors discuss “best practice” design patterns which are stated to positively influence the user retention of new users. However, none of the sources reviewed showcase any statistically significant proof of this claim. Thus, the objective of this study was to: Design and implement a set of commonly applied design patterns within a digital product’s user onboarding experience and evaluate their effects on user retention Through A/B testing on the SaaS product GetAccept, the following two design patterns were evaluated: Reduce friction – reducing the number of barriers and steps for a new user when first using a digital product; and Monitor progress – monitoring and clearly showcasing the progress of a new user’s journey when first using a digital product. The retention metric used to evaluate the two design patterns was first week user retention, defined as the share of customers who after signing up, sign in again at least once within one week. This was tested by randomly assigning new users into different groups: groups that did receive changes related to the design patterns, and one group did not receive any changes. By then comparing the first week user retention data between the groups using Fisher’s exact test, the conclusion could be drawn that with statistical significance, both of the evaluated design patterns positively influenced user retention for GetAccept. Furthermore, due to the generalizable nature of GetAccept’s product and the aspects evaluated, this conclusion should also be applicable to other companies and digital products with similar characteristics, and the method used to evaluate the impact of implementing the design patterns should be applicable for evaluating other design patterns and/or changes in digital products. However, as the method used for data collection in the study could not ensure full validity of it, the study could and should be repeated with the same design patterns on another digital product and set of users in order to strengthen the reliability of the conclusions drawn.

波松分配（Poisson Distribution）常用在單位時間或是區間內，計算對有興趣之某隨機事件次數（或是已知事件之頻率），例如：速食餐廳的單位時間來客數，又或是每段期間內，某天然災害的發生次數，可以表示為某一特定事件X服從波松分配，若lambda為單位事件發生次數或是平均次數，我們稱lambda為此波松分配之母數，記作Poisson(lambda)，其中lambda屬於實數。 今天我們若想要探討由兩個服從不同波松分配抽取的隨機變數，如下列所述：令X={(X1,X2)}為一集合，其中Xi為X(i,1),X(i,2),...,X(i,ni)~Poisson(lambda(i)),i=1,2。欲探討兩波松分配之均數是否相同或相差小於某個常數d時，考慮以下檢定:H0:lambda2-lambda1<=d與H0:lambda2-lambda1>d，對於此問題可以使用的檢定方法有Przyborwski和Wilenski(1940)提出的條件檢定（Conditional test,C-test）或K.Krishnamoorthy與Jessica Thomson(2002)提出的精確性檢定（Exact test,E-test），其中的精確性檢定為一個非條件檢定（Unconditional Test）；K.Krishnamoorthy與Jessica Thomson比較條件檢定與精確性檢定的p-value皆小於顯著水準(apha)，而精確性檢定的檢定力不亞於條件檢定，因此精確性檢定比條件檢定更適合上面所述之假設問題。 Roger L.Berger（1996）提出一個以信賴區間的p-value所建立的較強力檢定，而目前只用於檢定兩二項分配（Binomial Distribution）的機率參數p是否相同為例，然而Berger在文中提到，較強力檢定比非條件檢定有更好的檢定力，而且要求的計算時間較少，可以提升檢定的效率。 本篇論文我們希望在固定apha與d時檢定的問題，建立一個兩波松分配均數顯著水準為apha的較強力檢定。 利用Roger L.Berger與Dennis D.Boos(1994)提出以信賴區間的p-value方法，建立波松分配兩兩母均數差的較強力檢定；研究發現此較強力檢定與精確性檢定的p-value皆小於apha，然而我們的檢定的檢定力皆不亞於精確性檢定所計算得出的檢定力，然而其apha及虛無假設皆需要善加考慮以本篇研究來看，當檢定為單尾檢定時，若apha<0.01，我們的較強力檢定沒有辦法找到比精確性檢定更好地拒絕域，換言之，此時較強力檢定與精確性檢定的檢定力將會相等。 / Poisson Distribution is used to calculate the probability of a certain phenomenon which attracted by researcher. If we want to test two random variable in an experiment .Therefore ,let X={(X1,X2)} be independent samples ,respectively ,from Poisson distribution ,also X(i,1),X(i,2),...,X(i,ni)~Poisson(lambda(i)),i=1,2. The problem of interest here is to test: H0:lambda2-lambda1<=d and H0:lambda2-lambda1>d, where 0<apha<1/2 ,and let Y1 equals sum of X1 and Y2 equals sum of X2, where apha ,lambda,d be fixed. In this problem of hypothesis testing about two Poisson means is addressed by the conditional test.However ,the exact method of testing based on the test statistic considered in K.Krishnamoorthy,Jessica Thomson(2002) also commonly used. Roger L.Berger ,Dennis D.Boos(1994) give a new way to calculate p-value,which replace the old method ,called it a valid p-value .In 1996, Roger L.Berger used the new way to propose a new test for two parameter of binomial distribution which is more powerful than exact test. In the other hand, Roger L.Berger also explain the unconditional test is more suitable than the conditional test. In this paper,we propose a new method for two parameter of Poisson distribution which revise from Roger L.Berger’s method. The result we obtain that our new test is really get a much bigger rejection region.We found when the fixed increasing ,the set of more powerful test increasing, and when the fixed power increasing ,the required sample size decreasing.

This diploma thesis is focused on the analysis of small samples where it is not possible to obtain more data. It can be especially due to the capital intensity or time demandingness. Where the production have not a wherewithall for the realization more data or absence of the financial resources. Of course, analysis of small samples is very uncertain, because inferences are always encumbered with the level of uncertainty.

本論文主要是探討在二維度空間上二母體分佈是否一致。我們利用排列 (permutation)檢定方法來做比較, 並藉由費雪(Fisher)正確檢定方法的想法而提出重標記 (relabel)排列檢定方法或稱為費雪排列檢定法。 我們透過可交換性的特質證明它是正確 (exact) 的並且比 Syrjala (1996)所建議的排列檢定方法有更高的檢定力 (power)。 本論文另提出二個空間模型: spatial multinomial-relative-log-normal 模型 與 spatial Poisson-relative-log-normal 模型 來配適一般在漁業中常有的右斜長尾次數分佈並包含很多0 的空間資料。另外一般物種可能因天性或自然環境因素像食物、溫度等影響而有群聚行為發生, 這二個模型亦可描述出空間資料的群聚現象以做適當的推論。 / This thesis proposes the relabel (Fisher's) permutation test inspired by Fisher's exact test to compare between distributions of two (fishery) data sets locating on a two-dimensional lattice. We show that the permutation test given by Syrjala (1996} is not exact, but our relabel permutation test is exact and, additionally, more powerful. This thesis also studies two spatial models: the spatial multinomial-relative-log-normal model and the spatial Poisson-relative-log-normal model. Both models not only exhibit characteristics of skewness with a long right-hand tail and of high proportion of zero catches which usually appear in fishery data, but also have the ability to describe various types of aggregative behaviors.