Site-specific multiple linear regression models were developed for eight sites in Ohio—six in the Western Lake Erie Basin and two in northeast Ohio on inland reservoirs--to quickly predict action-level exceedances for a cyanotoxin, microcystin, in recreational and drinking waters used by the public. Real-time models include easily- or continuously-measured factors that do not require that a sample be collected. Real-time models are presented in two categories: (1) six models with continuous monitor data, and (2) three models with on-site measurements. Real-time models commonly included variables such as phycocyanin, pH, specific conductance, and streamflow or gage height. Many of the real-time factors were averages over time periods antecedent to the time the microcystin sample was collected, including water-quality data compiled from continuous monitors. Comprehensive models use a combination of discrete sample-based measurements and real-time factors. Comprehensive models were useful at some sites with lagged variables (< 2 weeks) for cyanobacterial toxin genes, dissolved nutrients, and (or) N to P ratios. Comprehensive models are presented in three categories: (1) three models with continuous monitor data and lagged comprehensive variables, (2) five models with no continuous monitor data and lagged comprehensive variables, and (3) one model with continuous monitor data and same-day comprehensive variables. Funding for this work was provided by the Ohio Water Development Authority and the U.S. Geological Survey Cooperative Water Program.

This data set contains example data for exploration of the theory of regression based regionalization. The 90th percentile of annual maximum streamflow is provided as an example response variable for 293 streamgages in the conterminous United States. Several explanatory variables are drawn from the GAGES-II data base in order to demonstrate how multiple linear regression is applied. Example scripts demonstrate how to collect the original streamflow data provided and how to recreate the figures from the associated Techniques and Methods chapter.

Summary : Fuel demand is shown to be influenced by fuel prices, people's income and motorization rates. We explore the effects of electric vehicle's rates in gasoline demand using this panel dataset.

Files : dataset.csv - Panel dimensions are the Brazilian state ( i ) and year ( t ). The other columns are: gasoline sales per capita (ln_Sg_pc), prices of gasoline (ln_Pg) and ethanol (ln_Pe) and their lags, motorization rates of combustion vehicles (ln_Mi_c) and electric vehicles (ln_Mi_e) and GDP per capita (ln_gdp_pc). All variables are all under the natural log function, since we use this to calculate demand elasticities in a regression model.

adjacency.csv - The adjacency matrix used in interaction with electric vehicles' motorization rates to calculate spatial effects. At first, it follows a binary adjacency formula: for each pair of states i and j, the cell (i, j) is 0 if the states are not adjacent and 1 if they are. Then, each row is normalized to have sum equal to one.

regression.do - Series of Stata commands used to estimate the regression models of our study. dataset.csv must be imported to work, see comment section.

dataset_predictions.xlsx - Based on the estimations from Stata, we use this excel file to make average predictions by year and by state. Also, by including years beyond the last panel sample, we also forecast the model into the future and evaluate the effects of different policies that influence gasoline prices (taxation) and EV motorization rates (electrification). This file is primarily used to create images, but can be used to further understand how the forecasting scenarios are set up.

Sources: Fuel prices and sales: ANP (https://www.gov.br/anp/en/access-information/what-is-anp/what-is-anp) State population, GDP and vehicle fleet: IBGE (https://www.ibge.gov.br/en/home-eng.html?lang=en-GB) State EV fleet: Anfavea (https://anfavea.com.br/en/site/anuarios/)

IntroductionIntimate Partner Violence (IPV) is a worldwide public health problem and major human and legal rights abuses of women. It affects the physical, sexual, and psychological aspects of the victims therefore, it requires complex and multifaceted interventions. Health providers are responsible for providing essential healthcare services for IPV victims. However, there is a lack of detailed information on whether or not health providers are ready to identify and manage IPV. Therefore, this study aimed to assess health providers’ readiness and associated factors in managing IPV in public health institutions at Hawassa, Ethiopia.MethodInstitutional based cross-sectional study was conducted through a simple random sample of 424 health providers. Data was collected with an anonymous questioners using physician Readiness to Manage Intimate Partner Violence Survey (PREMIS) tool. Linear regression analysis was used to examine relationships among variables. The strength of association was assessed by using unstandardized β with 95% CI.ResultsThe mean score of perceived provider’s readiness in managing IPV was 26.18± 6.69. Higher providers age and providers perceived knowledge had positive association with provider perceived readiness in managing IPV. Whereas not had IPV training, absence of a protocol for dealing with IPV management, and provider attitude had a negative association with provider perceived readiness in managing IPV.Conclusion and recommendationThis study reviled that health providers had limited perceived readiness to manage IPV. Provision of training for providers and develop protocol for IPV managements have an important role to improve providers readiness in the managements of IPV.

There are surveys that gather precise information on an outcome of interest, but measure continuous covariates by a discrete number of intervals, in which case the covariates are interval censored. For applications with a second independent dataset precisely measuring the covariates, but not the outcome, this paper introduces a semiparametrically efficient estimator for the coefficients in a linear regression model. The second sample serves to establish point identification. An empirical application investigating the relationship between income and body mass index illustrates the use of the estimator.

Source/Credit: Michael Grogan https://github.com/MGCodesandStats https://github.com/MGCodesandStats/datasets/blob/master/cars.csv

Sample dataset for regression analysis. Given 5 attributes (age, gender, miles driven per day, debt, and income) predict how much someone will spend on purchasing a car. All 5 of the input attributes have been scaled to be in 0 to 1 range. Training set has 723 training examples. Test set has 242 test examples.

This dataset will be used in an upcoming Galaxy Training Network tutorial (https://training.galaxyproject.org/training-material/topics/statistics/) on use of feedforward neural networks for regression analysis.

We include Stata syntax (dummy_dataset_create.do) that creates a panel dataset for negative binomial time series regression analyses, as described in our paper "Examining methodology to identify patterns of consulting in primary care for different groups of patients before a diagnosis of cancer: an exemplar applied to oesophagogastric cancer". We also include a sample dataset for clarity (dummy_dataset.dta), and a sample of that data in a spreadsheet (Appendix 2).

The variables contained therein are defined as follows:

case: binary variable for case or control status (takes a value of 0 for controls and 1 for cases).

patid: a unique patient identifier.

time_period: A count variable denoting the time period. In this example, 0 denotes 10 months before diagnosis with cancer, and 9 denotes the month of diagnosis with cancer,

ncons: number of consultations per month.

period0 to period9: 10 unique inflection point variables (one for each month before diagnosis). These are used to test which aggregation period includes the inflection point.

burden: binary variable denoting membership of one of two multimorbidity burden groups.

We also include two Stata do-files for analysing the consultation rate, stratified by burden group, using the Maximum likelihood method (1_menbregpaper.do and 2_menbregpaper_bs.do).

Note: In this example, for demonstration purposes we create a dataset for 10 months leading up to diagnosis. In the paper, we analyse 24 months before diagnosis. Here, we study consultation rates over time, but the method could be used to study any countable event, such as number of prescriptions.

The meta-learning method proposed in this paper addresses the issue of small-sample regression in the application of engineering data analysis, which is a highly promising direction for research. By integrating traditional regression models with optimization-based data augmentation from meta-learning, the proposed deep neural network demonstrates excellent performance in optimizing glass fiber reinforced plastic (GFRP) for wrapping concrete short columns. When compared with traditional regression models, such as Support Vector Regression (SVR), Gaussian Process Regression (GPR), and Radial Basis Function Neural Networks (RBFNN), the meta-learning method proposed here performs better in modeling small data samples. The success of this approach illustrates the potential of deep learning in dealing with limited amounts of data, offering new opportunities in the field of material data analysis.

This research is concerned with a statistical method that has recently become widespread in the international literature; although, it is still limited in Hungarian research. The method is geographically weighted regression (GWR), which is demonstrated through an example of its application. GWR is a local model that is founded on the basis of regression, prominently taking into consideration the geogra phical distance. Since it does not calculate the global relations of the whole data, but concentrates on the relationship of the dependent and independent variables locally within a determined search area, it allows consideration of the spatially varying processes. Simply, GWR is a developed version of the global regression model, since, through its use, it is possible to take into account the local features that are hidden by the global approach.

We propose an extensive framework for additive regression models for correlated functional responses, allowing for multiple partially nested or crossed functional random effects with flexible correlation structures for, for example, spatial, temporal, or longitudinal functional data. Additionally, our framework includes linear and nonlinear effects of functional and scalar covariates that may vary smoothly over the index of the functional response. It accommodates densely or sparsely observed functional responses and predictors which may be observed with additional error and includes both spline-based and functional principal component-based terms. Estimation and inference in this framework is based on standard additive mixed models, allowing us to take advantage of established methods and robust, flexible algorithms. We provide easy-to-use open source software in the pffr() function for the R package refund. Simulations show that the proposed method recovers relevant effects reliably, handles small sample sizes well, and also scales to larger datasets. Applications with spatially and longitudinally observed functional data demonstrate the flexibility in modeling and interpretability of results of our approach.

This article studies the influence of leading by example on organizational psychological ownership and job psychological ownership. This article further introduces the mediating mechanism of organizational identification and the regulating mechanism of Leader–member Exchange (LMX). This study investigated 312 personnel from eight property management enterprises in East, Northwest, Northeast, and central China. This study adopts a quantitative research method, using survey data of project managers, team leaders, and managers of Property management projects in China. The data were collected by questionnaire survey. In terms of data analysis, AMOS 21.0 software was used to conduct structural equation modeling (SEM) using the maximum likelihood method to test direct and indirect effects. SPSS 25.0 software was used to test the moderating effect by multilevel regression analysis with the maximum variance method. Use these two methods to analyze the whole theoretical framework. The results established all assumed relationships. In this article, leading by example, one of the important dimensions of empowering leadership is studied as a new leadership style, and the predictive effect of leading by example on organizational psychological ownership and job psychological ownership is verified. This finding further verifies the influence mechanism and boundary conditions of empowering leadership in different dimensions. It is found that organizational identification has different mediating effects on leading by example and organizational psychological ownership and job psychological ownership. The moderating effect of LMX also further indicates that under the influence of Confucian pan-family culture, the leader's exemplary behavior with higher authority has a stronger influence on employees' organizational identification, organizational psychological ownership, and job psychological ownership. Their relationship is deeply influenced by the culture of China's unique organizational Circle Culture.

This database studies the performance inconsistency on the biomass HHV ultimate analysis. The research null hypothesis is the consistency in the rank of a biomass HHV model. Fifteen biomass models are trained and tested in four datasets. In each dataset, the rank invariability of these 15 models indicates the performance consistency.

The database includes the datasets and source codes to analyze the performance consistency of the biomass HHV. These datasets are stored in tabular on an excel workbook. The source codes are the biomass HHV machine learning model through the MATLAB Objected Orient Program (OOP). These machine learning models consist of eight regressions, four supervised learnings, and three neural networks.

An excel workbook, "BiomassDataSetUltimate.xlsx," collects the research datasets in six worksheets. The first worksheet, "Ultimate," contains 908 HHV data from 20 pieces of literature. The names of the worksheet column indicate the elements of the ultimate analysis on a % dry basis. The HHV column refers to the higher heating value in MJ/kg. The following worksheet, "Full Residuals," backups the model testing's residuals based on the 20-fold cross-validations. The article (Kijkarncharoensin & Innet, 2021) verifies the performance consistency through these residuals. The other worksheets present the literature datasets implemented to train and test the model performance in many pieces of literature.

A file named "SourceCodeUltimate.rar" collects the MATLAB machine learning models implemented in the article. The list of the folders in this file is the class structure of the machine learning models. These classes extend the features of the original MATLAB's Statistics and Machine Learning Toolbox to support, e.g., the k-fold cross-validation. The MATLAB script, name "runStudyUltimate.m," is the article's main program to analyze the performance consistency of the biomass HHV model through the ultimate analysis. The script instantly loads the datasets from the excel workbook and automatically fits the biomass model through the OOP classes.

The first section of the MATLAB script generates the most accurate model by optimizing the model's higher parameters. It takes a few hours for the first run to train the machine learning model via the trial and error process. The trained models can be saved in MATLAB .mat file and loaded back to the MATLAB workspace. The remaining script, separated by the script section break, performs the residual analysis to inspect the performance consistency. Furthermore, the figure of the biomass data in the 3D scatter plot, and the box plots of the prediction residuals are exhibited. Finally, the interpretations of these results are examined in the author's article.

Reference : Kijkarncharoensin, A., & Innet, S. (2022). Performance inconsistency of the Biomass Higher Heating Value (HHV) Models derived from Ultimate Analysis [Manuscript in preparation]. University of the Thai Chamber of Commerce.

Analysis-ready tabular data from "Predicting spatial-temporal patterns of diet quality and large herbivore performance using satellite time series" in Ecological Applications, Kearney et al., 2021. Data is tabular data only, summarized to the pasture scale. Weight gain data for individual cattle and the STARFM-derived Landsat-MODIS fusion imagery can be made available upon request. Resources in this dataset:Resource Title: Metadata - CSV column names, units and descriptions. File Name: Kearney_et_al_ECOLAPPL_Patterns of herbivore - metada.docxResource Description: Column names, units and descriptions for all CSV files in this datasetResource Title: Fecal quality data. File Name: Kearney_etal2021_Patterns_of_herbivore_Data_FQ_cln.csvResource Description: Field-sampled fecal quality (CP = crude protein; DOM = digestible organic matter) data and phenology-related APAR metrics derived from 30 m daily Landsat-MODIS fusion satellite imagery. All data are paddock-scale averages and the paddock is the spatial scale of replication and week is the temporal scale of replication. Fecal samples were collected by USDA-ARS staff from 3-5 animals per paddock (10% - 25% of animals in each herd) weekly during each grazing season from 2014 to 2019 across 10 different paddocks at the Central Plains Experimental Range (CPER) near Nunn, CO. Samples were analyzed at the Grazingland Animal Nutrition Lab (GANlab, https://cnrit.tamu.edu/index.php/ganlab/) using near infrared spectroscopy (see Lyons & Stuth, 1992; Lyons, Stuth, & Angerer, 1995). Not every herd was sampled every week or every year, resulting in a total of 199 samples. Samples represent all available data at the CPER during the study period and were collected for different research and adaptive management objectives, but following the basic protocol described above. APAR metrics were derived from the paddock-scale APAR daily time series (all paddock pixels averaged daily to create a single paddock-scale time series). All APAR metrics are calculated for the week that corresponds to the week that fecal quality samples were collected in the field. See Section 2.2.4 of the corresponding manuscript for a complete description of the APAR metrics. Resource Title: Monthly ADG. File Name: Kearney_etal2021_Patterns_of_herbivore_Data_ADG_monthly_cln.csvResource Description: Monthly average daily gain (ADG) of cattle weights at the paddock scale and the three satellite-derived metrics used to build regression model to predict AD: crude protein (CP), digestible organic matter (DOM) and aboveground net herbaceous production (ANHP). Data table also includes stocking rate (animal units per hectare) used as an interaction term in the ADG regression model and all associated data to derive each of these variables (e.g., sampling start and end dates, 30 m daily Landsat-MODIS fusion satellite imagery-derived APAR metrics, cattle weights, etc.). We calculated paddock-scale average daily gain (ADG, kg hd-1 day-1) from 2000-2019 for yearlings weighed approximately every 28-days during the grazing season across 6 different paddocks with stocking densities of 0.08 – 0.27 animal units (AU) ha-1, where one AU is equivalent to a 454 kg animal. It is worth noting that AU’s change as a function of both the number of cattle within a paddock and the size of individual animals, the latter of which changes within a single grazing season. This becomes important to consider when using sub-seasonal weight data for fast-growing yearlings. For paddock-scale ADG, we first calculated ADG for each individual yearling as the difference between the weights obtained at the end and beginning of each period, divided by the number of days in each period, and then averaged for all individuals in the paddock. We excluded data from 2013 due to data collection inconsistencies. We note that most of the monthly weight data (97%) is from 3 paddocks where cattle were weighed every year, whereas in the other 3 paddocks, monthly weights were only measured during 2017-2019. Apart from the 2013 data, which were not comparable to data from other years, the data represents all available weight gain data for CPER to maximize spatial-temporal coverage and avoid potential bias from subjective decisions to subset the data. Data may have been collected for different projects at different times, but was collected in a consistent way. This resulted in 269 paddock-scale estimates of monthly ADG, with robust temporal, but limited spatial, coverage. CP and DOM were estimated from a random forest model trained from the five APAR metrics: rAPAR, dAPAR, tPeak, iAPAR and iAPAR-dry (see manuscript Section 2.3 for description). APAR metrics were derived from the paddock-scale APAR daily time series (all paddock pixels averaged daily to create a single paddock-scale time series). All APAR metrics are calculated as the average of the approximately 28-day period that corresponds to the ADG calculation. See Section 2.2.4 of the manuscript for a complete description of the APAR metrics. ANHP was estimated from a linear regression model developed by Gaffney et al. (2018) to calculate net aboveground herbaceous productivity (ANHP; kg ha-1) from iAPAR. We averaged the coefficients of 4 spatial models (2013-2016) developed by Gaffney et al. (2018), resulting in the following equation: ANHP = -26.47 + 2.07(iAPAR) We first calculated ANHP for each day of the grazing season at the paddock scale, and then took the average ANHP for the 28-day period. REFERENCES: Gaffney, R., Porensky, L. M., Gao, F., Irisarri, J. G., Durante, M., Derner, J. D., & Augustine, D. J. (2018). Using APAR to predict aboveground plant productivity in semi-aid rangelands: Spatial and temporal relationships differ. Remote Sensing, 10(9). doi: 10.3390/rs10091474 Resource Title: Season-long ADG. File Name: Kearney_etal2021_Patterns_of_herbivore_Data_ADG_seasonal_cln.csvResource Description: Season-long observed and model-predicted average daily gain (ADG) of cattle weights at the paddock scale. Also includes two variables used to analyze patterns in model residuals: percent sand content and season-long aboveground net herbaceous production (ANHP). We calculated observed paddock-scale ADG for the entire grazing season from 2010-2019 (excluding 2013 due to data collection inconsistencies) by averaging seasonal ADG of each yearling, determined as the difference between the end and starting weights divided by the number of days in the grazing season. This dataset was available for 40 paddocks spanning a range of soil types, plant communities, and topographic positions. Data may have been collected for different projects at different times, but was collected in a consistent way. We note that there was spatial overlap among a small number paddock boundaries across different years since some fence lines were moved in 2012 and 2014. Model-predicted paddock-scale ADG was derived using the monthly ADG regression model described in Sections 2.3.3 and 2.3.4. of the associated manuscript. In short, we predicted season-long cattle weight gains by first predicting daily weight gain for each day of the grazing season from the monthly regression model using a 28-day moving average of model inputs (CP, DOM and ANHP ). We calculated the final ADG for the entire grazing season as the average predicted ADG, starting 28-days into the growing season. Percent sand content was obtained as the paddock-scale average of POLARIS sand content in the upper 0-30 cm. ANHP was calculated on the last day of the grazing season fusing a linear regression model developed by Gaffney et al. (2018) to calculate net aboveground herbaceous productivity (ANHP; kg ha-1) from satellite-derived integrated absorbed photosynthetically active radiation (iAPAR) (see Section 3.1.2 of the associated manuscript). We averaged the coefficients of 4 spatial models (2013-2016) developed by Gaffney et al. (2018), resulting in the following equation: ANHP = -26.47 + 2.07(iAPAR) REFERENCES: Gaffney, R., Porensky, L. M., Gao, F., Irisarri, J. G., Durante, M., Derner, J. D., & Augustine, D. J. (2018). Using APAR to predict aboveground plant productivity in semi-aid rangelands: Spatial and temporal relationships differ. Remote Sensing, 10(9). doi: 10.3390/rs10091474

This data release contains the input-data files and R scripts associated with the analysis presented in [citation of manuscript]. The spatial extent of the data is the contiguous U.S. The input-data files include one comma separated value (csv) file of county-level data, and one csv file of city-level data. The county-level csv (“county_data.csv”) contains data for 3,109 counties. This data includes two measures of water use, descriptive information about each county, three grouping variables (climate region, urban class, and economic dependency), and contains 18 explanatory variables: proportion of population growth from 2000-2010, fraction of withdrawals from surface water, average daily water yield, mean annual maximum temperature from 1970-2010, 2005-2010 maximum temperature departure from the 40-year maximum, mean annual precipitation from 1970-2010, 2005-2010 mean precipitation departure from the 40-year mean, Gini income disparity index, percent of county population with at least some college education, Cook Partisan Voting Index, housing density, median household income, average number of people per household, median age of structures, percent of renters, percent of single family homes, percent apartments, and a numeric version of urban class. The city-level csv (city_data.csv) contains data for 83 cities. This data includes descriptive information for each city, water-use measures, one grouping variable (climate region), and 6 explanatory variables: type of water bill (increasing block rate, decreasing block rate, or uniform), average price of water bill, number of requirement-oriented water conservation policies, number of rebate-oriented water conservation policies, aridity index, and regional price parity. The R scripts construct fixed-effects and Bayesian Hierarchical regression models. The primary difference between these models relates to how they handle possible clustering in the observations that define unique water-use settings. Fixed-effects models address possible clustering in one of two ways. In a "fully pooled" fixed-effects model, any clustering by group is ignored, and a single, fixed estimate of the coefficient for each covariate is developed using all of the observations. Conversely, in an unpooled fixed-effects model, separate coefficient estimates are developed only using the observations in each group. A hierarchical model provides a compromise between these two extremes. Hierarchical models extend single-level regression to data with a nested structure, whereby the model parameters vary at different levels in the model, including a lower level that describes the actual data and an upper level that influences the values taken by parameters in the lower level. The county-level models were compared using the Watanabe-Akaike information criterion (WAIC) which is derived from the log pointwise predictive density of the models and can be shown to approximate out-of-sample predictive performance. All script files are intended to be used with R statistical software (R Core Team (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org) and Stan probabilistic modeling software (Stan Development Team. 2017. RStan: the R interface to Stan. R package version 2.16.2. http://mc-stan.org).

This material brings data on the results of univariate gamma regression model for direct costs, which was the first stage of inferential analysis using the linear regression model so that we could analyse which variables could be interacting with total direct cost per capita. The first table shows these data and precedes the multivariate analysis described in the article. The second table shows a more detailed descreptive analysis of per capita direct costs according to the current drug use pattern (evaluated by ASSIST alcohol, cannabis and cocaine/crack), including mean, standard deviation, minimum, maximum, first quartile, median, third quartile and the p value according to Kruskal-Wallis test. These data make reference to the article by Dr. Paula Becker e Dr. Denise Razzouk called " Relationships between age of onset of drug use, use pattern, and direct health costs in a sample of adults’ drug dependents in treatment at a Brazilian community mental health service ".

These files contain the data and data dictionary used to produce the results presented in: Moeltner, K., Puri, R., Johnston, R. J., Besedin, E., Balukas, J. A., & Le, A. (2023). Locally-weighted meta-regression and benefit transfer. Journal of Environmental Economics and Management, 121, 102871.

Meta-regression models (MRMs) are commonly used within benefit transfer to estimate broadly applicable, “umbrella” benefit functions that may be used to predict willingness to pay for environmental quality improvements at sites for which primary valuation studies have not been conducted. In virtually all benefit transfers of this type, a single regression model is fit to all source points in the metadata, and used to produce out-of-sample predictions for all possible policy-site applications. Despite the advantages of this approach over other types of benefit transfer, the predictive accuracy of these MRMs generally leaves room for improvement. This dataset enables reproduction of the presented locally-weighted regression approach to MRM estimation, for an empirical application on willingness-to-pay for water quality improvements. The metadata are drawn from primary stated preference studies that estimate per household (use and nonuse) WTP for water quality changes in specific U.S. water bodies. Changes in water quality, in turn, affect ecosystem services including aquatic life support, recreational uses (such as fishing, boating, and swimming), and nonuse values. Studies were limited to those for which WTP estimates could be readily mapped to water quality changes measured on a standard 100-point Water Quality Index (WQI). All monetary values were adjusted to 2019 U.S. dollars. The data includes 188 observations from 58 prior stated preference studies, with the earliest of these published in 1980. Variable definitions are provided in the attached data dictionary file. Additional details of the metadata are described in Moeltner et al. (2023).

This dataset contains the repository data used for our study "A Large-Scale Study of Modern Code Review and Security in Open Source Projects". This dataset was collected from GitHub, and includes 3,126 projects in 143 languages, with 489,038 issues and 382,771 pull requests. We also include the regression analysis notebooks for reproducing our results from this data.

Context

This file contains 5 years of daily time series data for several measures of traffic on a statistical forecasting teaching notes website whose alias is statforecasting.com. The variables have complex seasonality that is keyed to the day of the week and to the academic calendar. The patterns you you see here are similar in principle to what you would see in other daily data with day-of-week and time-of-year effects. Some good exercises are to develop a 1-day-ahead forecasting model, a 7-day ahead forecasting model, and an entire-next-week forecasting model (i.e., next 7 days) for unique visitors.

Content

The variables are daily counts of page loads, unique visitors, first-time visitors, and returning visitors to an academic teaching notes website. There are 2167 rows of data spanning the date range from September 14, 2014, to August 19, 2020. A visit is defined as a stream of hits on one or more pages on the site on a given day by the same user, as identified by IP address. Multiple individuals with a shared IP address (e.g., in a computer lab) are considered as a single user, so real users may be undercounted to some extent. A visit is classified as "unique" if a hit from the same IP address has not come within the last 6 hours. Returning visitors are identified by cookies if those are accepted. All others are classified as first-time visitors, so the count of unique visitors is the sum of the counts of returning and first-time visitors by definition. The data was collected through a traffic monitoring service known as StatCounter.

Inspiration

This file and a number of other sample datasets can also be found on the website of RegressIt, a free Excel add-in for linear and logistic regression which I originally developed for use in the course whose website generated the traffic data given here. If you use Excel to some extent as well as Python or R, you might want to try it out on this dataset.

These four labeled data sets are targeted at ordinal quantification. The goal of quantification is not to predict the label of each individual instance, but the distribution of labels in unlabeled sets of data.

With the scripts provided, you can extract CSV files from the UCI machine learning repository and from OpenML. The ordinal class labels stem from a binning of a continuous regression label.

We complement this data set with the indices of data items that appear in each sample of our evaluation. Hence, you can precisely replicate our samples by drawing the specified data items. The indices stem from two evaluation protocols that are well suited for ordinal quantification. To this end, each row in the files app_val_indices.csv, app_tst_indices.csv, app-oq_val_indices.csv, and app-oq_tst_indices.csv represents one sample.

Our first protocol is the artificial prevalence protocol (APP), where all possible distributions of labels are drawn with an equal probability. The second protocol, APP-OQ, is a variant thereof, where only the smoothest 20% of all APP samples are considered. This variant is targeted at ordinal quantification tasks, where classes are ordered and a similarity of neighboring classes can be assumed.

Usage

You can extract four CSV files through the provided script extract-oq.jl, which is conveniently wrapped in a Makefile. The Project.toml and Manifest.toml specify the Julia package dependencies, similar to a requirements file in Python.

Preliminaries: You have to have a working Julia installation. We have used Julia v1.6.5 in our experiments.

Data Extraction: In your terminal, you can call either

make

(recommended), or

julia --project="." --eval "using Pkg; Pkg.instantiate()"
julia --project="." extract-oq.jl

Outcome: The first row in each CSV file is the header. The first column, named "class_label", is the ordinal class.

Further Reading

Implementation of our experiments: https://github.com/mirkobunse/regularized-oq

Replication Data for: "Embedding Regression: Models for Context-Specific Description and Inference". All necessary data and estimated models are available in the following Dropbox folder: https://www.dropbox.com/sh/7al371qtr9102qq/AADKhjhYgnFCxOOQaugQloTBa?dl=0 Keep in mind the folder is quite large at 12.59 GB. Paper Abstract: Social scientists commonly seek to make statements about how word use varies over circumstances—including time, partisan identity, or some other document-level covari- ate. For example, researchers might wish to know how Republicans and Democrats diverge in their understanding of the term “immigration.” Building on the success of pretrained language models, we introduce the `a la Carte on Text (conText) embed- ding regression model for this purpose. This fast and simple method produces valid vector representations of how words are used—and thus what words “mean”—in dif- ferent contexts. We show that it outperforms slower, more complicated alternatives, and works well even with very few documents. The model also allows for hypothesis testing and statements about statistical significance. We demonstrate that it can be used for a broad range of important tasks, including understanding US polarization, historical legislative development, and sentiment detection. We provide open-source software for fitting the model.