120 likes | 346 Views
CSC 595 Lecture #5 Hypothesis & Argument formulation. Reference Book: Justin Zobel , Writing For Computer Science, 2 nd Ed. Springer, 2004. Dr. Manar Hosny. The Literature Review Model. Addresses. Step 3. Develop the Argument. Step 4. Survey the Literature. Step 6.
E N D
CSC 595Lecture #5Hypothesis & Argument formulation Reference Book: Justin Zobel, Writing For Computer Science, 2nd Ed. Springer, 2004 Dr. ManarHosny
The Literature Review Model Addresses Step 3. Develop the Argument Step 4. Survey the Literature Step 6. Write the Review Step 1. Select a topic Step 2. Search the Literature Step 5. Critique the Literature Advocates and defines Specifies and frames Documents and discovers Explores and catalogs Organizes and forms
Hypothesis • The first stages of a research involve identifying a specific issue to investigate develop a specific question to answer • For example, how something works, interacts, or behaves • This forms a hypothesis of your research • In computer science a hypothesis is usually about whether a proposed approach is fit for a certain purpose
Hypothesis • For example, a research hypothesis may be: “Using algorithm X we can reduce the number of memory accesses and make the program faster” • Another research hypothesis may be: “Sorting algorithm Y can be improved if we replace a tree-based structure with an array-based structure” • The goal of the research will be to test this hypothesis • The hypothesis should be clear, precise and unambiguous • Often is important to state what is not being proposed (the limitations of the hypothesis)
Hypothesis • Consider that there are two data structures: P-List, and Q-List and you believe that Q-list is superior to P-lists • Now consider the following hypotheses: • Q-lists are superior to P-lists • Wrong Hypothesis (un-testable) Success has to apply in all applications, in all conditions, for all times • As an in-memory search structure for large data-sets, Q-lists are faster and more compact than P-lists • Good Hypothesis (testable) the scope is limited to a domain that can feasibly be explored
Hypothesis • Another important factor is that the hypothesis must not be vague • Q-List performance is comparable to P-List performance • Vague Hypothesis • Our proposed query language is relatively easy to learn • Vague Hypothesis
Hypothesis • Sometimes you need to refine the hypothesis as a result of initial testing • However, this does not mean that the hypothesis should follow the experiment • The hypothesis makes a prediction • The hypothesis is confirmed if the prediction is successful • Tests should be blind • If the experiment has been fine-tuned to fit the hypothesis, we cannot say that the experiment confirmed the hypothesis
Defending Hypotheses • As part of the research process, you need to test your hypothesis and assemble supporting evidence. • The argument relates your hypothesis to the evidence. • In constructing an argument, imagine that you want to defend your hypothesis to a colleague and convince her that your hypothesis is correct. • Rebut likely objections • Admit when you are not certain • Anticipate problems that the reader may have with your hypothesis • Search for counter examples
Defending Hypotheses • For example if you want to defend the hypothesis that “algorithm X is faster than algorithm Y”, your argument can be: • Since the complexity of algorithm X is O(log2n) and the complexity of algorithm Y is O(n) algorithm X is faster than argument Y • A hypothesis may be wrong. • In that case, DO NOT cling to it and twist the results to defend it
Evidence • There are 4 types of evidence that can be used to support your hypothesis: • An analysis or proof (e.g. complexity analysis) • A model: a mathematical description of the hypothesis • A simulation: an implementation or partial implementation of simplified form of the hypothesis, done in a tightly controlled environment • An experiment: a full test of the hypothesis, based on real data