data scientists toolbox course notes

数据科学工具课程,包含命令行,R语言,Markdown等用法

CLI (命令行 Command Line Interface)

• / = root directory
• ~ = home directory
• pwd = print working directory (current directory)
• clear = clear screen
• ls = list stuff
  • -a = see all (hidden)
  • -l = details
• cd = change directory
• mkdir = make directory
• touch = creates an empty file
• cp = copy
  • cp <file> <directory> = copy a file to a directory
  • cp -r <directory> <newDirectory> = copy all documents from directory to new Directory * -r = recursive
• rm = remove
  • -r = remove entire directories (no undo)
• mv = move
  • move <file> <directory> = move file to directory
  • move <fileName> <newName> = rename file
• echo = print arguments you give/variables
• date = print current date

GitHub 代码仓库

• Workflow
  1. make edits in workspace
  2. update index/add files
  3. commit to local repo
  4. push to remote repository
• git add . = add all new files to be tracked
• git add -u = updates tracking for files that are renamed or deleted
• git add -A = both of the above
  • Note: add is performed before committing
• git commit -m "message" = commit the changes you want to be saved to the local copy
• git checkout -b branchname = create new branch
• git branch = tells you what branch you are on
• git checkout master = move back to the master branch
• git pull = merge you changes into other branch/repo (pull request, sent to owner of the repo)
• git push = commit local changes to remote (GitHub)

Markdown 文本编辑

• ## = signifies secondary heading (bold big font)
• ### = signifies tertiary heading (slightly smaller font than secondary, not bold)
• * = bullet list item

R Packages

• Primary location for R packages (rightarrow) CRAN
• available.packages() = all packages available
• head(rownames(a),3) = returns first three names of a
• install.packages("nameOfPackage") = install single package
• install.packages(c("nameOfPackage", "nameOfPackage", "nameOfPackage") = install multiple package
• Bioconductor Packages:
  • source("https://bioconductor.org/biocLite.R")
  • biocLite() = install bioconductor packages
• library(packagename) = load package
• search() = see all functions in package after loading

Types of Data Science Questions

• in order of difficulty: Descriptive (rightarrow) Exploratory (rightarrow) Inferential (rightarrow) Predictive (rightarrow) Causal (rightarrow) Mechanistic
• Descriptive analysis = describe set of data, interpret what you see (census, Google Ngram)
• Exploratory analysis = discovering connections (correlation does not = causation)
• Inferential analysis = use data conclusions from smaller population for the broader group
• Predictive analysis = use data on one object to predict values for another (if X predicts Y, does not = X cause Y)
• Causal analysis = how does changing one variable affect another, using randomized studies, Strong assumptions, golden standard for statistical analysis
• Mechanistic analysis = understand exact changes in variables in other variables, modeled by empirical equations (engineering/physics

Data

• Data = values of qualitative or quantitative variables, belonging to a set of items (usually population)
• Variables = measurement/characteristic of an item (qualitative vs quantitative)
• Data = not always structured, usually raw file, different formats
• Most important thing is question, then it is data
• Big data = now possible to collect data cheap, but not necessarily all useful (need the right data)

Experimental Design

• Formulate you question in advance
• Statistical inference = select subset, run experiment, calculate descriptive statistics, use inferential statistics to determine if results can be applied broadly
• [Inference] Variability = lower variability + clearer differences = decision
• [Inference] Confounding = underlying variable might be causing the correlation (sometimes called Spurious correlation)
  • dealing with confounding: fix variables, stratify (all options), randomize
• [Prediction] collection observations for different variable values, build predictive functions
  • similar problems of probability/sampling and confounding variables
• [Prediction] Difficult to understand where observation is from from different distributions. (size of effects important)
• [Prediction] Positive/negative statuses: True positive, false positive, false negative, true negative
  • Sensitivity = Pr(positive test | disease)
  • Specificity = Pr(negative test | no disease)
  • Positive Predictive Value = Pr(disease | positive test)
  • Negative Predictive Value = Pr(no disease | negative test)
  • Accuracy = Pr(correct outcome)
• Data dredging = use data to fit hypothesis
• Good experiments = have replication, measure variability, generalize problem, transparent
• Prediction is not inference, and be ware of data dredging