\relax 
\citation{Boehm1981}
\citation{Shepperd1996}
\citation{Auer2006}
\citation{Pendharkar2005}
\citation{Mendes2008a}
\citation{Li2009}
\citation{Menzies2010xx,koc2011}
\@writefile{toc}{\contentsline {section}{\numberline {1}Introduction}{1}}
\newlabel{sec:introduction}{{1}{1}}
\citation{baker07,Menzies2006}
\citation{Mendes2003,Auer2006,baker07,Li2009}
\citation{Boehm1981}
\citation{kemerer87}
\@writefile{toc}{\contentsline {section}{\numberline {2}Motivation}{2}}
\newlabel{sec:motivation}{{2}{2}}
\citation{Auer2006,Walkerden1999,Kirsopp2003,Shepperd1997,Shepperd1996,Kadoda2000,Li2008,Li2006,Li2007,Li2009,keung2008a,keung2008b,keung2008c}
\citation{Myrtveit}
\citation{Menzies2010xx}
\citation{Menzies2010xx}
\citation{Menzies2006}
\citation{koc2011}
\citation{Boehm1981}
\citation{Menzies2010xx}
\citation{koc2011jj}
\citation{Mendes2002,Mendes2003}
\citation{Briand1999,jeffery2001,Mendes2002,Mendes2003}
\citation{Palpanas2003,John1995,Frank03}
\citation{Browman1999}
\citation{Browman1999}
\citation{Mendes2002,Mendes2003}
\citation{Browman1999}
\citation{Browman1999}
\citation{Shepperd2007}
\citation{Jorgensen2004}
\citation{Boehm2000}
\@writefile{toc}{\contentsline {section}{\numberline {3}On the Value of Negative Results}{4}}
\newlabel{sec:negative-results}{{3}{4}}
\citation{Boehm1981}
\citation{Menzies2006,Kadoda2001}
\citation{Briand1997,Klas2008}
\citation{Trendowicz2006}
\citation{Shepperd1996}
\citation{Mendes2003}
\citation{Li2009}
\@writefile{toc}{\contentsline {section}{\numberline {4}Background}{5}}
\newlabel{sec:background}{{4}{5}}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.1}Software Effort Estimation}{5}}
\newlabel{subsec:software-effort-estimation}{{4.1}{5}}
\citation{Mendes2002a}
\citation{Briand1999}
\citation{Myrtveit1999}
\citation{Angelis2000}
\citation{Mendes2002a}
\citation{Angelis2000}
\citation{Mendes2003}
\citation{Wand1994}
\citation{Scheid2004}
\citation{Palpanas2003}
\citation{Frank03}
\citation{John1995}
\newlabel{equ:euclid}{{1}{6}}
\@writefile{toc}{\contentsline {subsection}{\numberline {4.2}Kernel Density Estimation}{6}}
\newlabel{subsec:kernel-estimation}{{4.2}{6}}
\citation{Mendes2002,Mendes2003}
\citation{Scott1992,Cressie1993}
\citation{Scheid2004,Wand1994}
\citation{John1995}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces We see a Gaussian kernel density estimate built on individual data points. Each point is in the center of a kernel and its effect is distributed to its neighborhood. The sum of all kernels make up the final Gaussian kernel density estimate.}}{7}}
\newlabel{fig:pointwise-kernel}{{1}{7}}
\newlabel{equ:general-kernel}{{2}{7}}
\newlabel{equ:general-kernel-b}{{3}{7}}
\newlabel{equ:general-kernel-g}{{4}{7}}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces The formulas for different kernels used in this study. In formulas $\rho = {\begingroup x - X_i\endgroup \over h}$. Note that IRWM kernel has different characteristics and its calculation details were provided in Section 4.1\hbox {}.}}{7}}
\newlabel{fig:kernel-formulas}{{2}{7}}
\citation{Boehm1981}
\citation{Desharnais1989}
\citation{Alpaydin2004}
\citation{Kadoda2000}
\@writefile{toc}{\contentsline {section}{\numberline {5}Methodology}{8}}
\newlabel{sec:methodology}{{5}{8}}
\@writefile{toc}{\contentsline {subsection}{\numberline {5.1}Weighting Method}{8}}
\newlabel{subsec:weighting-method}{{5.1}{8}}
\newlabel{equ:wkde}{{5}{8}}
\newlabel{equ:weight}{{6}{8}}
\newlabel{equ:effort-update}{{7}{8}}
\@writefile{toc}{\contentsline {subsubsection}{\numberline {5.1.1}Uniform vs. Non-Uniform Weighting}{8}}
\@writefile{toc}{\contentsline {subsection}{\numberline {5.2}Data}{8}}
\newlabel{fig:weighting-abe}{{5.1.1}{9}}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces In the case of ABE0 all instances are given equal probability values, hence equal weights. However, uniform kernel prefers some instances over the others: Only a certain portion of the instances are given equal non-zero weights.}}{9}}
\newlabel{fig:weight-vs-non-weight}{{3}{9}}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces We used 699 projects coming from 19 datasets. Datasets have different characteristics in terms of the number of attributes as well as the measures of these attributes.}}{9}}
\newlabel{fig:datasets}{{4}{9}}
\@writefile{toc}{\contentsline {subsection}{\numberline {5.3}Experiments}{9}}
\newlabel{sect:expr}{{5.3}{9}}
\citation{Molokken-Ostvold2004,Briand1999,Foss}
\citation{Stensrud2002}
\citation{Kitchenham2001}
\citation{Mendes2008a}
\citation{Shepperd1997}
\citation{Foss}
\@writefile{toc}{\contentsline {subsection}{\numberline {5.4}Performance Criteria}{10}}
\newlabel{ar}{{8}{10}}
\newlabel{EquationMRE}{{9}{10}}
\newlabel{EquationMdRE}{{10}{10}}
\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces Comparing algorithms (\textit  {i},\textit  {j}) on performance ($P_i$,$P_j$). The ``better'' predicate changes according to $P$. For error measures like MRE, ``better'' means lower medians. However, for PRED(25), ``better'' means higher medians.}}{11}}
\newlabel{FigureWinTieLossPseudocode}{{5}{11}}
\@writefile{toc}{\contentsline {section}{\numberline {6}Results}{11}}
\newlabel{sec:results}{{6}{11}}
\@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces Desharnais dataset $win,tie,loss$ statistics for ABE0 and N-ABE through Gaussian kernel. For each dataset we have $4$ of these tables (one for each kernel). In total it amounts to \textit  {19 Datasets $\times $ 4 tables = 76 tables}. In addition we have another \textit  {19 datasets $\times $ 1 kernel = 19 tables} from IRWM kernel. It is infeasible to include all the tables in this paper, therefore an executive summary of $76+19=95$ tables is provided in Figure\nobreakspace  {}7\hbox {}. Furthermore, we provide all $95$ tables in excel format at http://goo.gl/qpQiD.}}{12}}
\newlabel{fig:win-tie-loss-desharnais}{{6}{12}}
\@writefile{lof}{\contentsline {figure}{\numberline {7}{\ignorespaces Nine data sets comparing ABE0 to N-ABE. For every row in each cell, there are three symbols indicating the effect of N-ABE w.r.t. 3 different error measures. From left to right, the first symbol stands for N-ABE effect w.r.t. MdMRE, the second symbol w.r.t. MAR and the third one w.r.t. Pred(25). A ``$+$'' indicates that for majority of $k$ values (at least 3 out of 5 \textit  {k} values), N-ABE improved ABE0 in terms of $win-loss$ values. ``$-$'' indicates that N-ABE decreased the performance of ABE0 in the majority case. If the former conditions do not satisfy, then a ``$o$'' symbol is assigned. Note that dataset order here is the same as Figure\nobreakspace  {}4\hbox {}, yet the dataset names are abbreviated to 3 to 5 letters due to space constraints.}}{13}}
\newlabel{fig:summary-all1}{{7}{13}}
\@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces Ten more data sets comparing ABE0 to N-ABE. Same format as Figure\nobreakspace  {}7\hbox {}.}}{14}}
\newlabel{fig:summary-all2}{{8}{14}}
\@writefile{lof}{\contentsline {figure}{\numberline {9}{\ignorespaces The comparison of ABE0 to N-ABE under IRWM kernel. Similar to Figure\nobreakspace  {}7\hbox {} three symbols indicate the effect of N-ABE w.r.t. 3 different error measures and ``$+$'' indicates that for majority of $k$ values N-ABE improved ABE0 in terms of $win-loss$ values. A ``$-$'' symbol indicates a decrease and a ``$o$'' symbol indicates neither decrease nor increase. Notice that subject to IRWM kernel, N-ABE fails to improve ABE0 w.r.t. 3 different performance measures.}}{15}}
\newlabel{fig:summary-irwm}{{9}{15}}
\@writefile{toc}{\contentsline {section}{\numberline {7}Discussion}{15}}
\newlabel{sec:discussion}{{7}{15}}
\@writefile{lof}{\contentsline {figure}{\numberline {10}{\ignorespaces An intuitive example. In a 3 analogy case, there is a $75\%$ change for a hypothetical test project between uniform and non-uniform weighting.}}{15}}
\newlabel{fig:intuitive-example}{{10}{15}}
\citation{Palpanas2003,Frank03,John1995}
\citation{Alpaydin2004}
\citation{Milic2004}
\citation{Robson2002}
\citation{Kitchenham2001}
\citation{Kitchenham2009}
\@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {50 Sample Points: Note the bad fit due to low sample size.}}}{16}}
\@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {100 Sample Points: Note the better fit due to increased sample size.}}}{16}}
\@writefile{lof}{\contentsline {subfigure}{\numberline{(c)}{\ignorespaces {1000 Sample Points: Note the optimum fit due to high sample size.}}}{16}}
\@writefile{toc}{\contentsline {section}{\numberline {8}Threats to Validity}{16}}
\newlabel{sec:threats-to-validity}{{8}{16}}
\newlabel{fig:50-sample-points}{{11(a)}{17}}
\newlabel{sub@fig:50-sample-points}{{(a)}{17}}
\newlabel{fig:100-sample-points}{{11(b)}{17}}
\newlabel{sub@fig:100-sample-points}{{(b)}{17}}
\newlabel{fig:1000-sample-points}{{11(c)}{17}}
\newlabel{sub@fig:1000-sample-points}{{(c)}{17}}
\@writefile{lof}{\contentsline {figure}{\numberline {11}{\ignorespaces The effect of sample size and bandwidth on kernel density estimation. The choice of optimum bandwidth (\textit  {h} value) is important. However, even with the optimum bandwidth, one still needs enough number of samples for successful estimation. Sample size of $50$ appears to be too small and when we increase it to $100$, we get a better fit. Yet, for a very close fit, we need to go up to $1000$ sample points.}}{17}}
\newlabel{fig:sample-size-kernel-estimation}{{11}{17}}
\citation{Palpanas2003,John1995}
\@writefile{toc}{\contentsline {section}{\numberline {9}Conclusions}{18}}
\newlabel{sec:conclusions}{{9}{18}}
\@writefile{toc}{\contentsline {subsection}{\numberline {9.1}Answers To Research Questions}{18}}
\@writefile{toc}{\contentsline {section}{\numberline {10}Future Work}{18}}
\newlabel{sec:future-work}{{10}{18}}
\bibstyle{abbrv}
\bibdata{myref}
\bibcite{Alpaydin2004}{1}
\bibcite{Angelis2000}{2}
\bibcite{Auer2006}{3}
\bibcite{baker07}{4}
\bibcite{Boehm2000}{5}
\bibcite{Boehm1981}{6}
\bibcite{Briand1997}{7}
\bibcite{Briand1999}{8}
\bibcite{Browman1999}{9}
\bibcite{Cressie1993}{10}
\bibcite{Desharnais1989}{11}
\bibcite{Foss}{12}
\bibcite{Frank03}{13}
\bibcite{jeffery2001}{14}
\bibcite{John1995}{15}
\bibcite{Jorgensen2004}{16}
\bibcite{Jorgensen2007}{17}
\bibcite{Kadoda2000}{18}
\bibcite{kemerer87}{19}
\bibcite{keung2008a}{20}
\bibcite{keung2008c}{21}
\bibcite{koc2011jj}{22}
\bibcite{keung2008b}{23}
\bibcite{Kirsopp2003}{24}
\bibcite{Kitchenham2009}{25}
\bibcite{Kitchenham2001}{26}
\bibcite{Klas2008}{27}
\bibcite{koc2011}{28}
\bibcite{Li2006}{29}
\bibcite{Li2008}{30}
\bibcite{Li2007}{31}
\bibcite{Li2009}{32}
\bibcite{Mendes2002a}{33}
\bibcite{Mendes2008a}{34}
\bibcite{Mendes2002}{35}
\bibcite{Mendes2003}{36}
\bibcite{Menzies2006}{37}
\bibcite{Menzies2010xx}{38}
\bibcite{Milic2004}{39}
\bibcite{Molokken-Ostvold2004}{40}
\bibcite{Myrtveit1999}{41}
\bibcite{Myrtveit}{42}
\bibcite{Palpanas2003}{43}
\bibcite{Pendharkar2005}{44}
\bibcite{Robson2002}{45}
\bibcite{Scheid2004}{46}
\bibcite{Scott1992}{47}
\bibcite{Shepperd2007}{48}
\bibcite{Kadoda2001}{49}
\bibcite{Shepperd1997}{50}
\bibcite{Shepperd1996}{51}
\bibcite{Stensrud2002}{52}
\bibcite{Trendowicz2006}{53}
\bibcite{Walkerden1999}{54}
\bibcite{Wand1994}{55}
\citation{Jorgensen2007}
\citation{Browman1999}
\citation{Browman1999}
\citation{Mendes2002,Mendes2003}
\citation{Browman1999}
\citation{Browman1999}
\citation{Kitchenham2001}