R - Bayesian

 prop_model(data)

rbinom(n = 200 how many times to run the simulation, size = 100 sample size, prob = 0.42 probability)

runif(n = 6, min = 0.0, max = 1.0), random uniform sample between 0-1, 6 samples. continuous
dunif Discrete version

rbeta(n_draws, shape1 = 5, shape2 = 95), generate a distribution, the larger shape1,2 the more concentrated the distribution is, larger shape1 makes the distribution closer to 1, larger shape2 makes the distribution closer to 0

 

Poisson distribution

rpois(n_draws, lambda = mean_clicks)

 

Efficient alternative

dbinom(x = x1, size = sample size, prob= given probability) calculates specific probability, like P( x = 10 | p = 10%). args similar to rbinom though.

To generate a distribution instead of calculation one prob only, use x1 <- seq(0,100, by = 1) for example, or prob <- seq(0,1,by 0.01)

expand.grid(x,y) to generate all combination rows of 2 vectors

 

Normal distribution

x <- data

dnorm(x, mean = …, sd = …) calculate the likelihood for each datapoint in x. To create a distribution with dnorm, set x <- seq().

rnorm(n =  … , mean = …, sd = …) simulate a distribution with n samples

Deal with very small percentage probability with log probability

visialize with plot(x,y, type = “h”)

 

2 complete models, bino and normal

2018-11-08 18_05_20-A Bayesian model of water temperature _ R2018-11-08 18_05_30-A Bayesian model of water temperature _ R

BEST model

library(BEST) John Kurshner

BESTmcmc()

 

Comments

Post a Comment

Popular posts from this blog

Bryan Peterson – Learning to See Creatively

  Expanding your vision Try different, new angles, i.e lying and climbing to shoot Close up shooting tells better story and evoke stronger feelings. Elements of Design Line Shape Form Pattern Color Composition Zoom in, fill the frame Rule of Third Frame within frame, another object acts as a frame. Picture within picture, part of the current image can be an image on its own. Just take both horizontal and vertical pics or … not, just break the rules The Magic of Light Front lighting – no notes Sidelighting – most dramatic, contrast between highlights and shadows Backlighting – sihoulette, shadow Digital Photography Using photo retouching software is fine Career Considerations – did not read  
Read more

R – Stats, factor count,proportion

  Change to factor: adult$RACEHPR2 <- factor(adult$RACEHPR2, labels = c(“Latino”, “Asian”, “African American”, “White”)) show all factors: levels(v) droplevels(v) to remove unused factors Counts of factors: x <- table(adult$RBMI, adult$SRAGE_P) or count(dataframe, variable) Proportion: prop.table(x ) sum all = 1 prop.table(x,1 ) conditional proportions, sum row = 1 prop.table(x,1 ) conditional proportions, sum column = 1  
Read more

R - Supervised Learning

Image
  knn k nearest neighbors, k library(class) knn(training data, test data, labels), extra args: k = 2. prob = TRUE enables probability of each guess. Primitive bayes nrows(df), count rows subset(df, conditions/ col > 0 etc.), nrows this subset then do divisions Naive bayes model <- naive_bayes(y ~ x1 + x2 + … , data = df) y as explained by x1, x2 etc. predict(model , test_df) + args type = “prob” show posterior probability (instead of priori overall prob) Laplace correction:  add 1 to prevent the 0% probability affects all estimation Logistic Regression model <- glm(y ~ x1 + x2 …, data = df, family = “binomial”) predict(model, test_df, type = “response”) ROC curve to measure how well the model performs, AUC area under curve value from 0 – 1, 1 is best. library(pROC) ROC <- roc(actual var, predicted var) plot(ROC), auc(ROC) Decision Tree library(rpart) model <- rpart(y ~ x1 + x2 + …, data = df, method = “class”, control = rpart.control(cp = 0)) predict(m...
Read more