-
Notifications
You must be signed in to change notification settings - Fork 1
/
README.Rmd
177 lines (143 loc) · 4.93 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
---
output: github_document
---
<!-- README.md is generated from README.Rmd. Please edit that file -->
```{r, include = FALSE}
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>",
fig.path = "man/figures/README-",
out.width = "100%"
)
```
# dbmm: Dynamic Bayesian Measurement Models #
<!-- badges: start -->
<!-- badges: end -->
## Installation
You can install the development version of **dbmm** from [GitHub](https://github.com/) with:
``` r
# install.packages("devtools")
devtools::install_github("devincaughey/dbmm")
```
You will also need to install **cmdstanr** (for instructions, see [here](https://mc-stan.org/cmdstanr/articles/cmdstanr.html)).
## Overview ##
The R package **dbmm** fits dynamic Bayesian measurement models using the
programming language [Stan](https://mc-stan.org) via the R package [**cmdstanr**](https://mc-stan.org/cmdstanr/). Currently, the only supported model is a dynamic factor (DF) model for indicators of mixed type (binary, ordinal, or metric). In the future, however, the package will incorporate other models, including the dynamic group-level item response theory (DGIRT) model currently implemented by the R package [**dgo**](https://github.com/jamesdunham/dgo).
## Workflow ##
Using **dbmm** involves the following steps:
1. *Shape* the data into the list format required by **cmdstanr**.
2. *Fit* the required model in Stan using.
3. *Extract* parameter draws from the fitted model.
4. If needed, *identify* the model the model by rotating and/or sign-flipping the draws.
5. *Check* the convergence diagnostics of the fitted model.
5. *Label* the draws with informative parameter names.
6. *Summarize* and *plot* the posterior distributions.
### Step 1: Shape data
```{r, eval=FALSE}
## Load data on societal attributes of U.S. states in 2020 and 2021
data("social_outcomes_2020_2021")
## Drop observations with missing values
social_outcomes_2020_2021 <- na.omit(social_outcomes_2020_2021)
## Shape the data into list form
shaped_data <- shape_data(
long_data = social_outcomes_2020_2021,
unit_var = "st",
time_var = "year",
item_var = "outcome",
value_var = "value",
periods_to_estimate = 2020:2021,
ordinal_items = NA,
binary_items = NA,
max_cats = 10,
standardize = TRUE,
make_indicator_for_zeros = TRUE
)
```
### Step 2: Fit the model
You can specify additional arguments for `cmdstanr::sample()`. For details, see [here](https://mc-stan.org/cmdstanr/reference/model-method-sample.html).
```{r,eval=FALSE}
options(mc.cores = parallel::detectCores()) # for parallizing across chains
fitted <- fit(
data = shaped_data,
n_dim = 2,
chains = 2,
parallelize_within_chains = FALSE,
constant_alpha = FALSE,
separate_eta = TRUE,
init_kappa = FALSE,
force_recompile = FALSE,
iter_warmup = 500,
iter_sampling = 500,
adapt_delta = .9,
refresh = 10,
seed = 123
)
```
### Step 3: Extract draws ###
```{r,eval=FALSE}
fitted_draws <- extract_draws(fitted)
head(fitted_draws)
```
### Step 4: Identify the model ###
```{r,eval=FALSE}
identified_draws <- identify_draws(fitted_draws, rotate = TRUE)
## (To apply varimax rotation, set `rotate = TRUE`.)
```
### Step 5: Check convergence of the identified model ###
```{r,eval=FALSE}
## Basic check
check_convergence(identified_draws$id_draws)
## Traceplot of selected parameters
bayesplot::mcmc_trace(identified_draws$id_draws, pars = "lambda_metric[23,2]")
## More details
summarized_draws <- summary(identified_draws$id_draws)
summary(summarized_draws)
```
### Step 6: Label parameters ###
```{r,eval=FALSE}
labeled_draws <- label_draws(identified_draws)
head(labeled_draws$eta)
head(labeled_draws$lambda_metric)
```
### Step 7: Summarizing and plotting the posterior draws
```{r,eval=FALSE}
library(tidyverse)
## Posterior mean and standard deviation of the factor scores and item loadings
eta_summ <- labeled_draws$eta %>%
summarise(
est = mean(value),
err = sd(value),
.by = c(TIME, UNIT, dim)
)
head(eta_summ)
lambda_metric_summ <- labeled_draws$lambda_metric %>%
summarise(
est = mean(value),
err = sd(value),
.by = c(ITEM, dim)
)
head(lambda_metric_summ)
## Plot item loadings
lambda_metric_summ %>%
pivot_wider(
id_cols = "ITEM",
names_from = "dim",
values_from = c("est", "err")
) %>%
ggplot() +
aes(x = est_1, y = est_2, label = ITEM) +
geom_vline(xintercept = 0, linetype = "dotted") +
geom_hline(yintercept = 0, linetype = "dotted") +
geom_point() +
geom_linerange(
aes(xmin = est_1 - 1.96*err_1, xmax = est_1 + 1.96*err_1),
alpha = 1/4, linewidth = 2
) +
geom_linerange(
aes(ymin = est_2 - 1.96*err_2, ymax = est_2 + 1.96*err_2),
alpha = 1/4, linewidth = 2
) +
ggrepel::geom_text_repel() +
labs(title = "Item Loadings") +
coord_fixed()
```