CLIMR_exploration-past-data_report.Rmd

---
title: "Construal Level International Multilab Replication (CLIMR) Project: Analysis of BIF Response Option Valence Differences in Past Data -- Sánchez et al (2021)"
author: "CLIMR Team"
date: "`r Sys.Date()`"
output:
  html_document:
    theme: darkly
    toc: yes
    toc_float: yes
    keep_md: true
  pdf_document:
    toc: yes
knit: (function(input_file, encoding) {
    rmarkdown::render(input_file, rmarkdown::github_document(html_preview = FALSE), encoding = encoding, output_dir = "./reports/")
  })
---

# Re-Analysis of Sánchez et al (2021) BIF data

Given that we unexpectedly found a significant interaction between the distance
manipulation and BIF response option valence differences for the temporal
distance replication, it may be worthwhile to examine whether there is evidence
for the influence of this confound in previous data.

To explore this possibility, we fit a series of mixed effects logistic
regression models on the BIF item responses in the data of Sánchez et al (2021)
in which we examine a potential interaction between the temporal distance
manipulation in this study and valence differences in the item response options.

```{r}
lrt_val_sanchez
```

A likelihood ratio test indicates that adding the valence differences and the
interaction term significantly improves the model.

```{r}
summary(glmm_sanchez_bif_int)
```

Although Sánchez et al (2021) reported a significant effect for temporal
distance on the BIF, accounting for the interaction with the valence difference
confound, this effect becomes near-zero and nonsignificant. The interaction is
such that as the valence difference increases, the effect of the distance
manipulation increases in magnitude. Converting from the log odds scale to _d_,
the effect for the distance manipulation when the valence difference is zero is
_d_ = `r glmm_sanchez_d_report`.

This interaction is in a direction opposite to what was observed in the CLIMR
replication of the temporal distance effect.

## Models Not Retained

```{r}
summary(glmm_sanchez_bif_base)
```

```{r}
summary(glmm_sanchez_bif_val)
```

# Reanalysis of Grinfeld et al (2024, Study 2C) BIF data

We can also examine data from Grinfeld et al (2024, Study 2C) to see if a
similar interaction occurs.

```{r}
lrt_val_grinfeld
```

A likelihood ratio test indicates significant model improvement when adding
valence differences as a predictor but not the interaction term.

```{r}
summary(glmm_grinfeld_bif_val)
```

Controlling for valence differences, the effect of hypotheticality holds in this
dataset, though there is also evidence that people prefer the abstract option
more when it is positive.

## Models Not Retained

```{r}
summary(glmm_grinfeld_bif_base)
```

```{r}
summary(glmm_grinfeld_bif_int)
```