ARM Models Sorted by Type

Peter Li edited this page Aug 11, 2014 · 2 revisions

Below are the types of regression models that have been categorized and coded in Stan from ARM:

Single Level Models


Multi-Level Models



Single-Level Models

Linear Models

No Predictors

lm (y ~ 1)

One Predictor

  • earnings: linear model with one predictor
lm (earnings ~ height)
  • earnings: linear model with one predictor and natural log transformation
lm (log_earnings ~ height)
  • earnings: linear model with one predictor and log10 transformation
lm (log10_earnings ~ height)
  • electric: linear model with one predictor
lm (post_test ~ treatment)
  • grades: linear model with one predictor
lm (final ~ midterm)
  • kid_iq: linear model with one predictor
lm (kid_score ~ mom_hs)
  • kid_iq: linear model with one predictor
lm (kid_score ~ mom_iq)
lm (y ~ x)
  • sesame: linear model with one predictor
lm (watched ~ encouraged)
  • sesame: linear model with one predictor
lm (y ~ encouraged)
  • sesame: linear model with one predictor
lm (y ~ watched_hat)
lm (y ~ y_lag)
  • weight: centered linear model
lm (y ~ c_height + 1)

Multiple Predictors with No Interaction

  • congress: linear model with two predictors
lm (vote_88 ~ vote_86 + incumbency_88)
  • earnings: linear model with two predictors and natural log transformation
lm (log_earnings ~ height + male)
  • earnings: linear model with two predictors and log log transformation
lm (log_earnings ~ log_height + male)
  • earnings: linear model with two predictors and log transformation
lm (log_earn ~ height + male, subset=earn>0)
  • earnings2: mlinear model with eleven predictors
lm (earnings ~ interest + male + over65 + white + immig + educ_r + workmos + workhrs_top + any_ssi 
               + any_welfare + any_charity)
  • earnings: linear model with ten predictors
lm (earnings ~ male + over65 + white + immig + educ_r + workmos + workhrs_top + any_ssi + any_welfare 
               + any_charity)
lm (earnings ~ male + over65 + white + immig + educ_r + any_ssi + any_welfare + any_charity)
  • electric: linear model with two predictors
lm (post_test ~ pre_test + treatment)
  • ideo: linear model with two predictors
lm (score1 ~ party + x, subset=overlap)
  • ideo: linear model with two predictors
lm (score1 ~ party + x, subset=incs)
  • ideo: linear model with two predictors and reparamaterization
lm (score1 ~ party + I(z*(party==0)) + I(z*(party==1)), subset=incs)
  • kid_iq: linear model with two predictors
lm (kid_score ~ mom_hs + mom_iq)
  • kid_iq: linear model with two predictors
lm (ppvt ~ hs + afqt)
  • kid_iq: linear model with one factor
lm (kid_score ~ as.factor(mom_work))
  • mesquite: linear model with six predictors
lm (weight~ diam1 + diam2 + canopy_height + total_height + density + group)
  • mesquite: linear model with six predictors and log transformation
lm (log_weight~ log_diam1 + log_diam2 + log_canopy_height + log_total_height + log_density + group)
  • nes: linear model with eight predictors
lm (partyid7 ~ real_ideo + race_adj + age30_44 + age45_64 + age65up + educ1 + gender + income)
  • sesame: linear model with three predictors and one factor
lm (watched ~ encouraged + pretest + as.factor(site) + setting)
  • sesame: linear model with three predictors and one factor
lm (y ~ watched_hat + pretest + as.factor(site) + setting)
  • finite_populations: linear model with appropriate calculations for calculating the standard deviation of a finite population
lm (g ~ u_1 + u)

Multiple Predictors with Interaction

  • earnings: linear model with two predictors and interaction and natural log transformation
lm (log_earnings ~ height + male + height:male)
  • earnings: linear model with two predictors and interaction and natural log transformation centered using z-score
lm (log_earnings ~ z_height + male + z_height:male)
  • electric: linear model with two predictors and interaction
lm (post_test ~ pre_test + treatment + pre_test:treatment)
  • ideo: linear model with two predictors and interaction
lm (score1 ~ party + x + party:x, subset=incs)
  • kid_iq: linear model with two predictors and interaction
lm (kid_score ~ mom_hs + mom_iq + mom_hs:mom_iq)
  • kid_iq: linear model with two predictors and interaction centered using mean
lm (kid_score ~ c_mom_hs + c_mom_iq + c_mom_hs:c_mom_iq)
  • kid_iq: linear model with two predictors and interaction centered using conventional points
lm (kid_score ~ c2_mom_hs + c2_mom_iq + c2_mom_hs:c2_mom_iq)
  • kid_iq: linear model with two predictors and interaction centered using z-score
lm (kid_score ~ z_mom_hs + z_mom_iq + z_mom_hs:z_mom_iq)
  • mesquite: linear model with one transformed predictor and log transformation
lm (log_weight ~ log_canopy_volume)
  • mesquite: linear model with three predictors and three transformed predictors and log transformation
lm (log_weight ~ log_canopy_volume + log_canopy_area + log_canopy_shape + log_total_height + log_density + group)
  • mesquite: linear model with one predictor and two transformed predictors and log transformation
lm (log_weight ~ log_canopy_volume + log_canopy_area + group)
  • mesquite: linear model with two predictors and three transformed predictors and log transformation
lm (log_weight ~ log_canopy_volume + log_canopy_area + log_canopy_shape + log_total_height + group)

Logit Regression Models

One Predictor

  • nes: generalized linear model with logit link function and one predictor
glm (vote ~ income, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and one predictor
glm (switc ~ dist, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and one predictor
glm (switc ~ dist100, family=binomial(link="logit"))

Multiple Predictors with No Interaction

  • earnings: generalized linear model with logit link function and two predictors
glm (earn_pos ~ height + male, family=binomial(link="logit"))
  • earnings_pt1: logistic regression model with eight predictors
glm (earnings ~ male + over65 + white + immig + educ_r + any_ssi + any_welfare + any_charity,
     family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and three predictors
glm (switc ~ dist100 + arsenic + educ4, family=binomial(link="logit"))

Multiple Predictors with Interaction

  • wells: generalized linear model with logit link function and two predictors and interaction
glm (switc ~ dist100 + arsenic + dist100:arsenic, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function with two predictors and interaction centered using mean
glm (switc ~ c_dist100 + c_arsenic + c_dist100:c_arsenic, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and four predictors and interaction centered using mean
glm (switc ~ c_dist100 + c_arsenic + c_dist100:c_arsenic + assoc + educ4, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and three predictors and interaction centered using mean
glm (switc ~ c_dist100 + c_arsenic + c_dist100:c_arsenic + educ4, family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and three predictors and interaction centered using mean
glm (switc ~ c_dist100 + c_arsenic + c_educ4 + c_dist100:c_arsenic + c_dist100:c_educ4 + c_arsenic:c_educ4, 
     family=binomial(link="logit"))
  • wells: generalized linear model with logit link function with three predictors and interaction with log transform and centered using mean
glm (switc ~ c_dist100 + c_log_arsenic + c_educ4 + c_dist100:c_log_arsenic + c_dist100:c_educ4     
             + c_log_arsenic:c_educ4, 
     family=binomial(link="logit"))
  • wells: generalized linear model with logit link function with three predictors and interaction with log transform and centered using mean
glm (switc ~ dist100 + log_arsenic + educ4 + dist100:log_arsenic + dist100:educ4 + log_arsenic:educ4, 
     family=binomial(link="logit"))
  • wells: generalized linear model with logit link function and three predictors with interaction
glm (switc ~ dist100 + arsenic + educ4 + dist100:arsenic, family=binomial(link="logit"))

Other Generalized Linear Regression Models

Poisson

  • roaches: poisson regression model with exposure and three predictors
glm (y ~ roach1 + treatment + senior, family=poisson, offset=log(exposure2))

Probit

  • wells: generalized linear model with probit link function and one predictor
glm (switc ~ dist100, family=binomial(link="probit"))

Quasipoisson

glm(y ~ roach1 + treatment + senior, family=quasipoisson, offset=log(exposure2))

Multi-Level Models

Linear Models

Varying Intercept

  • 8_schools: multi-level linear model with redundant parameterization

  • electric_1b: multi-level linear model with varying intercept and slope

lmer (y ~ treatment + pre_test + (1 | pair))
  • electric: multi-level linear model with varying intercept
lmer (y ~ treatment + (1 | pair))
lmer (y ~ 1 + (1 | county))
lmer (y ~ x + (1 | county))
  • radon_group: multi-level linear model with group level predictor and individual level predictors
lmer (y ~ x + u + (1 | county))
  • radon_redundant: multi-level liner model with varying intercept and redundant parameterization
lmer (y ~ 1 + (1 | county))
lmer (y ~ x + (1 | county))
lmer (y ~ x + (1 | county))
  • anova_radon_nopred: multi-level linear model with varying intercept and set up for ANOVA
lmer (y ~ 1 + (1 | county))
lmer (y ~ u + x + (1 | county))
lmer (y ~ u + x + x_mean + (1 | county))
lmer (y ~ u + (1 | county))
lmer (y ~ theta + (theta | j))

Varying Intercept and Slope

lmer (y ~ x (1 + x | ethn))
  • electric_1a: multi-level linear model with varying intercept and slope
lmer (y ~ 1 + (1 | pair) + (treatment | grade))
  • hiv: multi-level linear model with varying slope and intercept
lmer (y ~ time + (1 + time | person)
  • hiv_inter: multi-level linear model with interaction and varying slope and intercept
lmer (y ~ time:treatment + (1 + time | person)
  • radon_vary_si: multi-level linear model with group level predictors
lmer (y ~ 1 + (1 + x | county))
lmer (y ~ x + u.full + x:u.full + (1 + x | county))
  • r_sqr: multi-level linear model with appropriate calculations for R^2
lmer (y ~ 1 + (1 + x | county))

Multiple Group-Level Factors

lmer (y ~ x.centered + (1 + x.centered | eth) + (1 + x.centered | age) + (1 + x.centered | eth:age))
  • electric_1c: multi-level linear model with group level factors
lmer (y ~ 1 + (1 | pair) + (treatment | grade) + (pre_test | grade))
  • pilots: non-nested multi-level linear model with group level predictors
lmer (y ~ 1 + (1 | group.id) + (1 | scenario.id))
  • pilots: multi-level linear model with varying intercept and redundant parameterization
lmer (y ~ 1 + (1 | treatment) + (1 | airport))
  • pilots_expansion: multi-level linear model with varying intercept and parameter expansion
lmer (y ~ 1 + (1 | treatment) + (1 | airport))

Other

Logistic Regression Models

Varying Intercept

  • election88: multi-level logistic regression model with group level predictors
glmer (y ~ black + female + (1 | state), family=binomial(link="logit"))

Multiple Group-Level Factors

glmer (y ~ black + female + black:female + v.prev.full + (1 | age) + (1 | edu) + (1 | age.edu) 
          + (1 | state) + (1 | region.full), family=binomial(link="logit"))
  • election88: multi-level logistic regression model with redundant parameterization
glmer (y ~ female + black + female:black + (1 | age) + (1 | edu) + (1 | age_edu) + (1 | state), 
      family=binomial(link="logit"))
glmer (y ~ female + black + female:black + (1 | age) + (1 | edu) + (1 | age_edu) + (1 | state), 
      family=binomial(link="logit"))
  • item_response: multi-level logistic regression model with parameter expansion
glmer (y ~ a:g + (a:g | k,j) + (g:b | k), family=binomial(link="logit"))

Other Generalized Linear Regression Models

Bernoulli

  • dogs_log: multi-level model using binomial distribution

  • dogs_check: multi-level model using binomial distribution

Categorical

Multivariate Normal

Poisson

Robit

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.