FDA Finals Reviewer PDF

FDA FINALS REVIEWER Frequency Distribution – number of occurrences of an event Object – the representative or the one who is observing Relative Frequencies – percenta...

FDA FINALS REVIEWER Frequency Distribution – number of occurrences of an event Object – the representative or the one who is observing Relative Frequencies – percentages or proportion Attributes – characteristics/observations Cumulative Frequencies – value of Variables – attribute that is organized and use values that precede for analysis Measures of Association – quantify strength & Nominal Variable – distinctness, non-numerical direction of relationship Ordinal Variable – pertains to order Covariance – linear rs, change in one Interval Variable – not necessary start with zero can change another Ratio Variable – absolute zero origins Correlation – or Pearson correlation coefficient or “r”, both strength and linear rs Discrete – integer Measures of Shape – how data distributed, Continuous – real number pattern Data Sets Symmetric Distribution – mirror image Record Dataset – rows Asymmetric Distribution – biased, with M x N Data Matrix – fixed set numeric specific characteristics attributes Skewness – measure of asymmetry Document Data – vector Positive Skewed/Right-skewed/Positive Transaction Data – transaction, set of skewness – long right tail, positive mean, items median, mode Graph Data Set – interconnections Negative skewed/left-skewed/Negative Skewness – long left tail, negative mean, Ordered Data Set – sequence/pattern median, mode Population – entire collection of individuals Kurtosis – quantifies the shape of probability Parameters – characteristic/describe population distribution Sample – representatives from population Peakedness – data values are concentrated Statistics – describe a sample Mesokurtic – normal distribution Descriptive Statistics – describe & summarize Leptokurtic/heady-tailed distribution – more information than normal distribution, bell shaped Measures of Location – specified amount of Platykurtic/short-tailed distribution – less than percentage data. Deciles (10 parts). Percentiles normal (100 parts). Quartiles (4 parts) Inferential statistics – draw a conclusion Special Percentile – median (50th, half). SAMPLING First Quartile (25th, ¼). Second Quartile (50th=median) Law of Large Number – sample means get closer to the population mean as sample size Measures of Central Tendency – position of gets larger values identify the center, central tendency, mean, median, mode, average Central Limit Theorem – large sample will always fall to normal distribution Measures of Spread/Measures of Dispersion – how close or far, how spread Sampling Frame – Actual list of individuals Range – highest to lowest Probability Sampling – equal chance of being selected Variance – square of standard deviation Simple Random Sampling – everyone Standard Deviation – measures how has an opportunity spread out numbers Stratified Sampling – heterogenous to homogenous strata, propotionally  Cluster Sampling – divide into multi- DATA PRE-PROCESSING group or cluster Data Pre-Processing – improving quality for Systematic Sampling – random point to secondary analysis fixed interval DATA CLEANING Multistage Sampling – similar to cluster, Data Cleaning – address noise in data to ensure combination of other sampling accuracy and correctness Non-Probability Sampling – not everyone has a IMPORTANCE chance Ensures Data Accuracy and Reliability Quota Sampling – specific characteristic Improves Data Quality Purposive Sampling – specific criteria or purpose Reduces Errors and Bias in Analysis Snowball of Chain Sampling – referrals Supports Effective Decision-Making Convenience Sampling – availability DATA QUALITY ISSUES DESCRIPTIVE STATISTICS IN BUSINESS Missing Values – incomplete Understanding customer base – demographics Noise & Outliers – data that deviate Financial Analysis – profitability, liquidity Inconsistencies – error, formatting Market Research – trends, preferences Duplicate Data – repeated values Data Visualization – foundation to visualization DATA CLEANING PROCESS DESCRIPTIVE STATISTICS IN HR Handling Missing Values – imputing data Performance Management Smoothing Noisy Data – eliminating outliers Salary & Compensation Analysis Detecting and Deleting Outliers – use box plot Workforce Productivity Fix Structural Errors – all error in words Recruitment & Selection Remove Duplicates – avoid redundancy Evaluating HR Intervention Data Validation – authenticate your data INFERENTIAL STATISTICS IN BUSINESS TASKS A/B Testing – compare website who perform Filling in missing values better Ignore the tuple/row – exclude record Customer Satisfaction Survey – how satisfied with missing values, not effective if there’s a lot Retail Analytics – crucial for marketing & Filling in the missing value manually decisions Data Imputation – use global constant, INFERENTIAL STATISTICS IN HR attribute mean, attribute mean for all samples of the same class Predictive Analytics – develop predictive models Clean Noisy Data Resource Allocation & Optimization – predicting future outcomes Binning – categorical into numerical components Risk Assessment and Mitigation – identify & mitigate Clustering – use cluster average to represent value Hypothesis Testing – effectiveness, impact Regression – simple regression line to estimate very erratic dataset Combine Computer & Human Inspection – Human Interventions Box Plot – use to identify outliers DATA INTEGRATION DATA TRANSFORMATION TECHNIQUES Data Integration – combining data derived from Data Smoothing – helps predicting trends or various sources. Single data store patterns DATA INTEGRATION PROCESS Data Generalization – low-level to high-level Data Sour Identification Data extraction Attribution Construction – create new features Data Mapping Data Discretization – continuous data to data interval Data Validation & Quality Assurance Data Aggregation – compiling large volume then Data Transformation Data Loading transform Data Synchronization DATA TRANSFORMATION PROCESS Data Governance & Security Data Discovery Data Mapping Metadata Enhancement Data Extraction Code Generation & Execution Data access & analysis Review Sending TECHNIQUES IMPORTANCE Extract, Transform, Load (ETL) – secondary Standardization – consistency & uniformity in processing over, structured data, slower, high data cost, custom solutions Compatibility – integration of data for Extract, Load, Transform – loads directly, all comprehensive analysis types (structured, unstructured), faster, low cost, granular access control Analysis – easier interpretation and analysis Issues During Data Integration – redundant Visualization – easier to visualize data, data value conflicts, entity identification Data Profiling – understand the characteristics problem and quality of data FOUR TYPES OF DATA INTEGRATION: Clear Documentation – ensure repeatability and Inner Join – matching values quality Left Join – left values and matching Quality Assurance – ensure accuracy and values from the right integrity of data Right Join – right values and matching Automation – streamline and standardized values from the left processes Outer Join – the union of all values TECHNIQUES IMPORTANCE & USE Normalization – Adjusting data value Improved Decision-Making – Talent Standardization – Scaling data Management Encoding – converting from categorical Compliance with Regulations – Employee to numerical Engagement Aggregation – Combining multiple data Enhance Employee Insights – Workforce points Planning Filtering – Removing redundant or Streamlined Processes – Compensation & irrelevant data Benefits Imputation – Filling in missing values DATA TRANSFORMATION using mean, median, mode Data Transformation – is a process of OPTION FOR DATA TRANSFORMATION transforming data into another format suitable Normalization – scale specific variable falls to for the analysis. small specific range Min-Max Normalization – transform into a standardized format  Z-Score Standardization – transform Feature Creation – create new attributes that into a standard normal distribution captures important information Binning – transforming numerical to categorical feature creation methodologies Equal-width (distance) – uniform grid Feature Extraction – creating new features Equal-depth (frequency) – contains same number of samples Mapping Data or New Space – lower- dimensional space to higher-dimensional space DATA ENCODING – categorical to numerical Feature Construction – built Binary Encoding – transformation taking intermediate features values from 0 to 1 DATA POST-PROCESSING & VISUALIZATION Class-based Encoding – probability of the class variable Data Post-Processing – refining and preparing data for analysis after it has been collected. For Continuous Class – average of the class further analysis variable STEPS DATA REDUCTION Data Cleaning Data Reduction – reduced representation of data but produced same results Data Transformation DATA REDUCTION TECHNIQUES Data Integration Feature Selection – most relevant Data Reduction Dimensionality Reduction – preserve essential Data Enrichment – adding additional information features to enhance dataset’s value Data Compression – encode in more compact Data Validation – ensure that data meets required standard Inner Join – matching values This improves data quality, decision-making, Sampling and facilitates predictive analytics STRATEGIES Data Visualization – representing data in visual Sampling – smaller representative that R or R Studio – software facilities generalize the population Can do graphics, dashboards, matrix Simple Random Sampling – equal algebra, data handling & storage: numeric, probability textual, high-level data analytic & Statistical Sampling without replacement – each functions, and programming language: loops, item is selected, removed from the population branching, subroutine Sampling with replacement – not Can’t do: Not a database but connects removed to DBMS, no spreadsheet view of data, but connect to excel/ms offices, no Stratified Sampling – several partitions professional/commercial support Feature Subset Selection – reduces Strengths – Free and open source, strong user dimensionality, removing redundant community, highly extensible and flexible, Brute Force Approach – all possible flexible graphics and extensible defaults, feature subsets implementation of high-end statistical methods Embedded Approaches – feature Weaknesses – steep learning curve, and slow selection occurs naturally for very large datasets Filter Approaches – selected before PARTS data mining Source Pane – type/edit Wrapper Approaches – as black box to command/code find the best subset Console Pane – command line interface Environment Pane – displays object  Files Pane – acts as file explorer Bar Graph – a single point. Quantitative single point to qualitative single point WORKING WITH DATASET WAGE Scatter Graph/Plot – determine position. Data Frame – storing dataset, list of vectors (ex: Individual data points to two-dimensional point dataset wage) Pie Chart – dividing circle into proportional STRUCTURE OF DATASET slices Dataset – organized/ordered collection of data Histogram (numeric) – distribution or frequency Plotting the dataset – creating visuals of occurrence representation Box Plot – distribution through quartiles, BASIC VISUALIZATION OR WAYS TO PLOT DATA parallel Histogram – breaks data into bins (breaks) Tree Maps – hierarchal structures using rectangle size Scatter Plot – for simple data inspection Choropleth Maps – geographical areas, colored Box Plot – visualizing the spread and derive or patterns inferences Chart – structures that relate entities IN THE CONTEXT OF HR Network Diagrams – use nodes/vertices Recruitment & Selection – effectiveness of channels & attracting candidates Flow Chart – sequential, intricacies and decision Employee Retention – predict employee Diagrams – schematic picture, symbolic turnover Pareto Chart – identify prevalent defects Employment Engagement – enhance employee Side-by-side Chart – parallel chart, multiple data satisfaction series Training & Development – identify skill gaps & Steam-and-Leaf Display – organize and present training needs numerical data, identify patterns and outliers Plotting Inside R VISUAL DESIGN PRINCIPLES High-level functions – step-by-step Preattention – without the need of attention manner Color Selection Preattention (color) Low-level functions – detailed control Shape Selection Preattention (form) Geom_bar – describes values on the y axis (bar plot) Preattentive – takes a fixed amount of time Aes () function – specify desired aesthetics Conjunction Features DATA VISUALIZATION & COMMUNICATION Emergent Features, have or don’t have a unique feature Visualization – presentation of info using spatial or graphical representations Asymmetric Pre-attentive Features, asymmetric, sloped, and vertical line Comparison facilitation Block Text – specify desired aesthetics Recognition of patterns Visual Illusions – people don’t perceive the way General decision making they should Two types of using visualization, to Delboeuf illusion explore/calculate and to communicate Mueller-Lyon MEDIUM/TYPES OF VISUALIZATION Ebbinghaus illusion Graph – communicate information with at least two scales, symmetric paired Basic Categories Visual Representation – Graph, Table, Maps, Diagram, Network, Icon Line Graph – interval or a time span, trends & relationships CLASSIFICATION OF VISUAL REPRESENTATION Graphs – position and magnitude of geometric Graphical Tables – quantitative, rows & columns IMPLICATION TO DESCRIPTIVE ANALYTICS Tables – words, number, signs, combination Data Privacy Regulations – how info should be collected, used, and stored. Time Chart – temporal data Data Security – protect from unauthorized Network Chart – rs among components access Structure Diagrams – static description of physic Data Bias & Discrimination – prohibit using Process Diagram- interrelationships & processes biased data Maps – physical geography Data Subject Rights – grant individual rights Cartograms – spatial maps THREE C’s Icons – impart interpretation Consent Control Confidentiality Photorealistic Pictures – realistic pictures Credibility Character TUFTE’S GRAPHIC EXCELLENCE Multifunctional graphical elements TUFTE’S GRAPHICAL INTEGRITY Accurately represent data TUFTE’S INK MAXIMATION Maximize data-ink, minimize non-data- ink ETHICS Code of Ethics – principles of conduct Doing the right thing, protect research participants, ensure credibility, trust, accountability, reduce liabilities and waste resources 3 MAJOR ETHICAL CONCERN Discrimination & bias, integrity, lack of transparency Privacy – protection of personal information APPLICATION TO DESCRIPTIVE ANALYTICS Ethical Data Collection Avoiding Bias Honest Data Presentation Accountability and Impact Integrity – conducted honestly, no bias Transparency – openness and clarity Accountability – taking responsibility HANDLING DATA ETHICALLY Consent Data Minimization Security Access & Control Fair use Secure Disposal