Merge pull request #54 from NREL/laurenklun/doc_update

Update 3_sampling.tex

documentation/reference_doc/3_sampling.tex

@@ -49,7 +49,7 @@ \subsection{Building Type Assignments}
 \item[Hospital] Hospitals, wherever possible, are disambiguated from outpatient clinics through the existence of around-the-clock medical facilities. This is not possible in many states, in which case the differentiation is based on available CoStar data.
 \item[Large Hotel] Large hotels are differentiated from small hotels on the basis of conference or casino spaces. Hotels that have major facilities for conferences, events, or gambling are classified as large hotels.
 \item[Offices] Offices are divided up into three subsets: small, medium, and large. Each type of office is based on the thresholds used by American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) Appendix G \citep{ashrae_901_2010}, which include both size and number of stories. In the case of large offices, there are additional probability distributions that determine what percent (if any) of the office is a data center.
-\item[Outpatient] Outpatient facilities, as represented in ComStock, include non-hospital medical centers, rehabilitation centers, medical offices, and skilled nursing facilities.
+\item[Outpatient] Outpatient facilities, as represented in ComStock, include non-hospital medical centers, rehabilitation centers, and medical offices.
 \item[Primary School] The primary school type is used to represent all schools that do not include secondary or post-secondary education, i.e., grades 9 and beyond. Schools that provide education for pre-secondary to post-secondary students (e.g., grades 5--12) are classified as secondary schools. This grouping means that any daycare facilities classified as schools by HIFLD are included as primary schools, unless the facilities also support secondary students.
 \item[Quick Service Restaurant] Quick service restaurants consist entirely of fast food restaurants.
 \item[Retail] This category predominantly features large national retailers, excluding grocery stores. This includes big box stores, garden centers, department stores, and any other freestanding retailers that do not include a significant grocery section.
@@ -93,4 +93,4 @@ \section{Sampling Methodology}
 
 Several different classes of algorithms support sampling in multidimensional probability spaces. ComStock's choice has largely been influenced by \cite{sobol_sampling}, which introduced Sobol Sequences as a low-discrepancy basis for sampling in higher-dimensional spaces. This approach attempts, with each sample, to optimally reduce the maximal space between samples. Although this approach can be implemented on an iterative basis, ComStock adopted an approach developed by \cite{sobol_lib}, which uses a bit-switching algorithm to select optimal sampling within an $n$-dimensional probability space. This approach is designed to accurately capture the full extent of the distributions with limited samples on a non-biased basis. In other words, all buildings within a type are equally weighted after sampling, rather than being reweighted as a secondary step to sampling.
 
-The results of applying these algorithms produced 350,000 building samples. The set of characteristics for each sample defines the inputs to the ComStock BEM workflow, which creates a BEM for each sample. The characteristics of the 350,000 samples are recorded in the buildstock.csv file. The buildings described in this file provide the ComStock team's best estimate of the characteristics for a large portion of the United States' commercial building stock for use by researchers, practitioners, and consultants. For a detailed look at the accuracy of the resulting models, refer to \cite{eulp_final_report}. The building energy models for ComStock's 350,000 baseline building samples are available to download at \url{https://data.openei.org/} in the nrel-pds-building-stock data lake. See the README.md file for details.
+The results of applying these algorithms produced 350,000 building samples. The set of characteristics for each sample defines the inputs to the ComStock BEM workflow, which creates a BEM for each sample. The characteristics of the 350,000 samples are recorded in the buildstock.csv file. The buildings described in this file provide the ComStock team's best estimate of the characteristics for a large portion of the United States' commercial building stock for use by researchers, practitioners, and consultants. For a detailed look at the accuracy of the resulting models, refer to \cite{eulp_final_report}. The building energy models for ComStock's 350,000 baseline building samples are available to download at \url{https://data.openei.org/} in the nrel-pds-building-stock data lake. See the README.md file for details.