diff --git a/code/3lc_nocov.R b/code/3lc_nocov.R
index 4a780c87f9ff6eccef3509a4526ea8e221cb7736..3708e3224378b6b596a5df993f7eccee22bfeaf1 100644
--- a/code/3lc_nocov.R
+++ b/code/3lc_nocov.R
@@ -180,8 +180,8 @@ lc3cl[["unconditionals"]] <- apollo_lcUnconditionals(lc3cl,apollo_probabilities,
apollo_saveOutput(lc3cl, saveOutput_settings = list(saveEst=FALSE, savbmetv=FALSE, savbmetrr=FALSE))
+lc3lc_model <- choiceTools::quicktexregapollo(lc3cl)
-
-
+saveRDS(lc3lc_model, file = "modeloutput/lcmodelstex.RDS")
diff --git a/code/readmodels.R b/code/readmodels.R
index 833514eb49baaf743bc8f3ff4dfc23730ffe925f..2e53bf6caf9e3699f1d030d7b24ce01badb76108 100644
--- a/code/readmodels.R
+++ b/code/readmodels.R
@@ -1,9 +1,42 @@
-deleteold <-list.files("modeloutput/clogit/", pattern = "OLD", full.names=TRUE)
+deleteold <-list.files("modeloutput", pattern = "OLD", full.names=TRUE , recursive = TRUE)
file.remove(deleteold)
mnlsplits<-readRDS("modeloutput/modelstex.RDS")
-lc3cl_model <- readRDS("~/share/groups/bioecon/Julian/seminarheuristics/modeloutput/lclogit/lc3cl_model.rds")
\ No newline at end of file
+lc3cl_model <- readRDS("modeloutput/lcmodelstex.RDS")
+
+remGOF
+
+
+
+class1 <- lc3cl_model
+class2 <- lc3cl_model
+class3 <- lc3cl_model
+
+
+ for (ele in c("coef.names","coef","se","pvalues")) {
+ elements<- slot(class1,ele)[1:7]
+ slot(class1,ele) <- elements
+
+}
+
+
+for (ele in c("coef.names","coef","se","pvalues")) {
+ elements<- slot(class2,ele)[9:15]
+ slot(class2,ele) <- elements
+
+}
+
+for (ele in c("coef.names","coef","se","pvalues")) {
+ elements<- slot(class3,ele)[17:23]
+ slot(class3,ele) <- elements
+
+}
+
+allclasses <- remGOF(list(class2, class3))
+
+allclasses[["class1"]] <-class1
+
diff --git a/modeloutput/lcmodelstex.RDS b/modeloutput/lcmodelstex.RDS
new file mode 100644
index 0000000000000000000000000000000000000000..e12f561e193b65e03ed5f63e7f97f3ec19849da3
Binary files /dev/null and b/modeloutput/lcmodelstex.RDS differ
diff --git a/oldslides.R b/oldslides.R
deleted file mode 100644
index 31c76c2af265d04937ddb93653f6271022673bc1..0000000000000000000000000000000000000000
--- a/oldslides.R
+++ /dev/null
@@ -1,64 +0,0 @@
-## Example data set
-
-- We use a choice experiment with three attributes and two alternatives. There is no opt-out option. The design is a simple full factorial.
-
-## Attributes
-<div class=".hscroll">
- +-------------------------------+------------------+---------------------------------+
- | Attribute | Levels | Coding |
- +===============================+==================+=================================+
- | Country of Tofu manufacturing | Non EU | 0 0 |
- | | | |
- | | EU | 1 0 |
- | | | |
- | | Sweden | 0 1 |
- +-------------------------------+------------------+---------------------------------+
- | Country of origin of soybean | Non EU | 0 0 |
- | | | |
- | | EU | 1 0 |
- | | | |
- | | Sweden | 0 1 |
- +-------------------------------+------------------+---------------------------------+
- | Cultivation of Soybean | Conventional | 0 |
- | | | |
- | | Organic | 1 |
- +-------------------------------+------------------+---------------------------------+
- | Price per 400g in SEK | 15 SEK to 60 SEK | Continuous with original values |
- +-------------------------------+------------------+---------------------------------+
-
-
- ## Attributes
-
- ```{r attributes}
- df <- data.frame(
- Attribute = c(
- "Country of tofu manufacturing",
- "Country of origin of soybean",
- "Cultivation of Soybean",
- "Price per 400g in SEK"
- ),
- Levels = c(
- "Non EU | EU | Sweden",
- "Non EU | EU | Sweden",
- "Conventional | Organic",
- "15 SEK to 60 SEK"
- ),
- Coding = c(
- "0 | Dummy",
- "0 | Dummy",
- "0 | 1",
- "Continuous with original values"
- )
- )
-
-
- # Print the data frame using kable
- kable(df) %>% kable_styling(full_width = TRUE, position = "center")
-
- ```
-
-
- ## Example Choice Set
-
- 
-
\ No newline at end of file
diff --git a/slides.html b/slides.html
index b6ce1cdda838d79ab91c21d2b0429d805b4bb330..170806cf66f7adcc2387911152a3120c7032a452 100644
--- a/slides.html
+++ b/slides.html
@@ -344,9 +344,11 @@
<li><a href="#/data-generating-processes">Data generating processes</a></li>
<li><a href="#/example-feed-additives">Example: Feed Additives</a></li>
<li><a href="#/simulated-dataset">Simulated dataset</a></li>
-<li><a href="#/heuristics">Heuristics</a></li>
<li><a href="#/regression-results-with-all-models">Regression results with all models</a></li>
<li><a href="#/latent-class-models">Latent Class models</a></li>
+<li><a href="#/summary">Summary</a></li>
+<li><a href="#/way-forward">Way forward</a></li>
+<li><a href="#/now-it-is-your-turn">Now it is your turn</a></li>
</ul>
</nav>
</section>
@@ -7062,7 +7064,7 @@ Log Likelihood (Converged)
</div>
</div>
</section>
-<section id="heuristics" class="slide level2">
+<section id="heuristics" class="slide level2 unnumbered unlisted smaller">
<h2>Heuristics</h2>
<ul>
<li><p>In the data there are three different DGPs</p></li>
@@ -7076,7 +7078,7 @@ Log Likelihood (Converged)
<p><span class="math display">\[ U = \alpha + \beta_{cow}\text{cow}+\beta_{adv}\text{adv} + \beta_{vet}\text{vet} +
\beta_{far}\text{far}+\beta_{bon}\text{bon} + \epsilon\]</span></p>
<ul>
-<li>Group 3 (20%): Only Bonus with $ _{bon2}=1.9$</li>
+<li>Group 3 (20%): Only Bonus with $ _{bon2}=1.9 $</li>
</ul>
<p><span class="math display">\[ U = \beta_{bon2}\text{bon} + \epsilon\]</span></p>
</section>
@@ -7095,15 +7097,12 @@ Log Likelihood (Converged)
<li><p><a href="./choicesets.html">All respondents</a></p></li>
<li><p><a href="./choicesets_rum.html">RUM</a></p></li>
<li><p><a href="./choicesets_nonattendence.html">Attribute Non-Attendence</a></p></li>
-<li><p><a href="./choicesets_priceheuristic.html">Price Heuristics</a></p></li>
+<li><p><a href="./choicesets_priceheuristic.html">Bonus Heuristics</a></p></li>
</ul>
</section>
<section id="regression-results-with-all-models" class="slide level2 smaller">
<h2>Regression results with all models</h2>
<table class="texreg" style="margin: 10px auto;border-collapse: collapse;border-spacing: 0px;caption-side: bottom;color: #000000;border-top: 2px solid #000000;">
-<caption>
-Statistical models
-</caption>
<thead>
<tr>
<th style="padding-left: 5px;padding-right: 5px;">
@@ -7317,7 +7316,214 @@ Log Likelihood (Converged)
<section id="latent-class-models" class="slide level2">
<h2>Latent Class models</h2>
<ul>
-<li>We may try a Latent Class model to detect heuristics</li>
+<li><p>We may try a latent class model to detect heuristics</p></li>
+<li><p>In the ideal case, we would find two classes of about 20 % and one class with about 60%</p></li>
+<li><p>The classes would reflect the DGP in the three groups</p></li>
+</ul>
+</section>
+<section id="regression-results-latent-class" class="slide level2 unnumbered unlisted smaller">
+<h2>Regression results Latent Class</h2>
+<table class="texreg" style="margin: 10px auto;border-collapse: collapse;border-spacing: 0px;caption-side: bottom;color: #000000;border-top: 2px solid #000000;">
+<thead>
+<tr>
+<th style="padding-left: 5px;padding-right: 5px;">
+
+</th>
+<th style="padding-left: 5px;padding-right: 5px;">
+Class 1: 15%
+</th>
+<th style="padding-left: 5px;padding-right: 5px;">
+Class 2: 20%
+</th>
+<th style="padding-left: 5px;padding-right: 5px;">
+Class 3: 65%
+</th>
+</tr>
+</thead>
+<tbody>
+<tr style="border-top: 1px solid #000000;">
+<td style="padding-left: 5px;padding-right: 5px;">
+basc
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.35 (0.13)<sup>**</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-17.01 (9.81)<sup>*</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-0.03 (0.01)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+bcow
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.09 (0.06)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-0.21 (0.07)<sup>**</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.37 (0.03)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+badv
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-0.31 (0.21)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-15.47 (9.75)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.31 (0.03)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+bvet
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-0.23 (0.16)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-7.75 (7.61)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.15 (0.02)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+bfar
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-0.23 (0.34)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-7.75 (3.83)<sup>*</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.14 (0.02)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+bmet
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.66 (0.14)<sup>***</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+8.21 (4.85)<sup>*</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.15 (0.03)<sup>***</sup>
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+bbon
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.33 (0.03)<sup>***</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+18.04 (9.72)<sup>*</sup>
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+0.29 (0.01)<sup>***</sup>
+</td>
+</tr>
+<tr style="border-top: 1px solid #000000;">
+<td style="padding-left: 5px;padding-right: 5px;">
+No Observations
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+28800
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+No Respondents
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+3600
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+</tr>
+<tr>
+<td style="padding-left: 5px;padding-right: 5px;">
+Log Likelihood (Null)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-31640.03
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+</tr>
+<tr style="border-bottom: 2px solid #000000;">
+<td style="padding-left: 5px;padding-right: 5px;">
+Log Likelihood (Converged)
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+-19918.67
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+<td style="padding-left: 5px;padding-right: 5px;">
+
+</td>
+</tr>
+</tbody>
+</table>
+</section>
+<section id="summary" class="slide level2">
+<h2>Summary</h2>
+<ul>
+<li><p>Heuristics are likely present in the DGP</p></li>
+<li><p>Presence of heuristics violates standard RUM modelling</p></li>
+<li><p>Not accounting for heuristics can lead to biases</p></li>
+<li><p>How severe the biases are depends</p></li>
+<li><p>Latent class models are not necessarily useful</p></li>
+</ul>
+</section>
+<section id="way-forward" class="slide level2">
+<h2>Way forward</h2>
+<ul>
+<li><p>Check the data manually</p></li>
+<li><p>Try to identify respondents who apply some standard heuristics</p></li>
+<li><p>Identifying respondents who only regard one attribute may be easy</p></li>
+<li><p>Attribute Non attendence models may help as well (see tomorrow)</p></li>
+</ul>
+</section>
+<section id="now-it-is-your-turn" class="slide level2">
+<h2>Now it is your turn</h2>
+<ul>
+<li><p>You find the dataset <code>feedadd_data_practice.RDS</code>in the folder <code>data</code></p></li>
+<li><p>Have a look at the data and try to identify heuristics by either looking at choice frequencies or estimated models</p></li>
+<li><p>Estimate a four class LC model</p></li>
+<li><p>Try to write code that identifies a Bonus heuristic</p></li>
</ul>
<img src="iDiv_Logo_long.png" class="slide-logo r-stretch"><div class="footer footer-default">
diff --git a/slides.qmd b/slides.qmd
index aa23e884fec766566deeee392b31aae9b14d90bb..149643dd8ef129a83c953efab80f84273c58a775 100644
--- a/slides.qmd
+++ b/slides.qmd
@@ -43,7 +43,6 @@ source("code/readmodels.R")
-
- Often we assume a data generating process that aligns with the standard models we use for analysis
- Random utility model with Gumbel iid errors
@@ -178,7 +177,7 @@ kable(head(feedadd_data,n = 200)) %>% kable_styling(bootstrap_options = "striped
scroll_box(height = "600px", width = "100%")
```
-## Heuristics
+## Heuristics {.unnumbered .unlisted .smaller}
- In the data there are three different DGPs
@@ -192,7 +191,7 @@ $$ U = \alpha + \beta_{cow}\text{cow}+\beta_{adv}\text{adv} + \beta_{vet}\text{v
$$ U = \alpha + \beta_{cow}\text{cow}+\beta_{adv}\text{adv} + \beta_{vet}\text{vet} +
\beta_{far}\text{far}+\beta_{bon}\text{bon} + \epsilon$$
-- Group 3 (20%): Only Bonus with $ \beta\_{bon2}=1.9$
+- Group 3 (20%): Only Bonus with $ \beta\_{bon2}=1.9 $
$$ U = \beta_{bon2}\text{bon} + \epsilon$$
@@ -214,7 +213,7 @@ $$ U = \beta_{bon2}\text{bon} + \epsilon$$
- [Attribute Non-Attendence](./choicesets_nonattendence.html)
-- [Price Heuristics](./choicesets_priceheuristic.html)
+- [Bonus Heuristics](./choicesets_priceheuristic.html)
## Regression results with all models {.smaller}
@@ -227,4 +226,49 @@ texreg::htmlreg(mnlsplits, single.row = TRUE , custom.note = "" , html.tag = FAL
- We may try a latent class model to detect heuristics
+- In the ideal case, we would find two classes of about 20 % and one class with about 60%
+
+- The classes would reflect the DGP in the three groups
+
+
+
+
+## Regression results Latent Class{.unnumbered .unlisted .smaller}
+
+```{r lcreg, results='asis'}
+texreg::htmlreg(list(allclasses[["class1"]], allclasses[[1]],allclasses[[2]]), single.row = TRUE , custom.note = "" , html.tag = FALSE, head.tag = FALSE , body.tag = FALSE, caption = "", custom.model.names = paste0("Class ", 1:3 ,c(": 15%",": 20%",": 65%" )))
+```
+
+
+## Summary
+
+- Heuristics are likely present in the DGP
+
+- Presence of heuristics violates standard RUM modelling
+
+- Not accounting for heuristics can lead to biases
+
+- How severe the biases are depends
+
+- Latent class models are not necessarily useful
+
+## Way forward
+
+- Check the data manually
+
+- Try to identify respondents who apply some standard heuristics
+
+- Identifying respondents who only regard one attribute may be easy
+
+- Attribute Non attendence models may help as well (see tomorrow)
+
+## Now it is your turn
+
+- You find the dataset `feedadd_data_practice.RDS `in the folder `data`
+
+- Have a look at the data and try to identify heuristics by either looking at choice frequencies or estimated models
+
+- Estimate a four class LC model
+
+- Try to write code that identifies a Bonus heuristic