ARM Models Sorted by Type

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

Single Level Models

• Linear Models
  • No Predictors
  • One Predictor
  • Multiple Predictors with No Interaction
  • Multiple Predictors with Interaction
• Logit Regression Models
  • One Predictor
  • Multiple Predictors with No Interaction
  • Multiple Predictors with Interaction
• Other Generalized Linear Regression Models
  • Poisson
  • Probit
  • Quasipoisson

---

Multi-Level Models

• Linear Models
  • Varying Intercept
  • Varying Intercept and Slope
  • Multiple Group-Level Factors
• Logistic Regression Models
  • Varying Intercept
  • Multiple Group-Level Factors
  • Other
• Other Generalized Linear Regression Models
  • Bernoulli
  • Categorical
  • Multivariate Normal
  • Poisson
  • Robit

---

---

Single-Level Models

Linear Models

No Predictors

• lightspeed: linear model with 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)

• radon_complete_pool: multi-level linear model with complete pooling

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)

• unemployment: linear model with one predictor

lm (y ~ y_lag)

• weight: centered linear model

lm (y ~ c_height + 1)

• weight_censored: censored weight model

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)

• earnings_pt2: linear model with eight predictors

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

• roaches_overdispersion: poisson overdispersion regression model with exposure and three predictors

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))

• radon_intercept: multi-level linear model with varying intercept

lmer (y ~ 1 + (1 | county))

• radon_no_pool: multi-level linear model with no pooling

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))

• radon_vary_intercept_a: multi-level linear model with varying intercept set up to calculate pooling factors

lmer (y ~ x + (1 | county))

• radon_vary_intercept_b: multi-level linear model with varying intercept set up to calculate pooling factors

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))

• radon_vary_intercept_floor: multi-level linear model with varying intercept

lmer (y ~ u + x + (1 | county))

• radon_vary_intercept_floor2: multi-level linear model with varying intercept

lmer (y ~ u + x + x_mean + (1 | county))

• radon_vary_intercept_nofloor: multi-level linear model with varying intercept

lmer (y ~ u + (1 | county))

• multiple_comparisons: multi-level linear model that serves as a multiple comparisons example

lmer (y ~ theta + (theta | j))

Varying Intercept and Slope

• 17.1_radon_vary_inter_slope: varying intercept and slope model

• 17.1_radon_correlation: varying intercept and slope model with correlation between slopes and intercepts

• 17.2_radon_vary_inter_slope: varying intercept and slope model with group level predictors

• 17.2_radon_correlation: varying intercept and slope model with correlation between slopes and intercepts and group level predictors

• earnings_vary_si: multi-level linear model with group level predictors

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))

• radon_inter_vary: multi-level linear model with group level predictors

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

• earnings_latin_square: non-nested multi-level linear model with group level predictors

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

• 17.1_radon_wishart: scaled inverse wishart model

• 17.1_radon_wishart2: two varying coefficients model with unmodeled individual-level coefficients

• 17.2_radon_wishart: scaled inverse wishart model with group level predictors

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

• 17.4_multilevel_logistic: multilevel logistic regression model

• 17.7_latent_glm: latent-data parameterization of multilevel logistic regression model

• dogs: multi-level logit regression model

• election88_full: multi-level logistic regression model with group level predictors

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"))

• election88_expansion: multi-level logistic regression model with parameter expansion

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

• 17.6_multilevel_ordered_categorical: multilevel ordered categorical regression model

Multivariate Normal

• sesame_street1: multi-level linear model using multivariate normal

• sesame_street2: multi-level linear model using multivariate normal

• 17.1_radon_multi_varying_coef: multiply varying coefficients model

• 17.2_radon_multi_varying_coef: multiply varying coefficients model with group level predictors

Poisson

• 17.5_multilevel_poisson: multilevel poisson regression model

Robit

• 17.7_robit: robit regression model