Know your "standard outline of paper"

examples:

Intro

What is the problem? ideally, mentioning some recent paper that this paper uses as a starting point.
Why is it interesting and important?
Why is it hard? (E.g., why do naive approaches fail?)
Why hasn't it been solved before? (Or, what's wrong with previous proposed solutions? How does mine differ?)
What are the key components of my approach and results? Also include any specific limitations.

Research questions. You need 2+.

The contributions of this work are...

XXXX
XXXX

This work differs from our prior publications in this area [listed] since....

The rest of this article is structured as follows....

Background

Motivation.

Tell. A. Really. Good. Story. Why. This. Is. Important.

Related work.

Not what you are doing to do... but what everyone else is done. Presented respectfully, but critically. Blah blah did this but did not try that.

Methods

What data sets. What experimental rig. What statistical methods (hypothesis test and effect size test. when in doubt scott knott, bootstrap, cliff's delta). How you will report results.

Results

Graphics. Clear. Few in number. results from multipel samples presented in a small space.

RQ1: reults

RQ2: results. etc

etc

Threats to Validity

e.g. As with any empirical study, biases can affect the final results. Therefore, any conclusions made from this work must be considered with the following issues in mind:

Sampling bias threatens any classification experiment; i.e., what matters there may not be true here. For example, the data sets used here comes from the PROMISE repository and were supplied by one individual. Also even though we use ten open-source data sets for CPDP (Table III) and seven to run LACE (Table III), and the data covers a large scope of applications including text/xml processing systems, search engines, source code integration/build tools, and management information systems, they are all from Java systems.
Learner bias: For building the defect predictors in this study, we elected to use k-Nearest Neighbor. We chose the kNearest Neighbor because its results were comparable to the more complicated algorithms [38] and can act as a baseline for other algorithms. Classification is a large and active field and any single study can only use a small subset of the known classification algorithms.
Evaluation bias: This paper uses one measure of privacy, IPR. Other privacy measures used in software engineering include guessing anonymity [39], [40], and entropy [41], [42](discussed in §IV-D). Measuring privacy with other measures is left for future work.
Order bias: With LACE2, the order that the data owners get access to the private cache affects the amount of data that they submit to the cache. To mitigate this order bias, we run the experiment 10 times randomly changing the order of the data owners each time.
Input bias: For the MORPH algorithm, we randomly select input values for a set range to determine the boundary between the an instance and its nearest unlike neighbor within which we create MORPHed instances. Since different input values can result in different outputs, we mitigate this bias with 10 runs of the experiment for LACE1 and LACE2.
Conclusion, Future work

...

ai-se / admin