Aronson APLS Analyses

Hypotheses Overview

H1a: Administrators who are motivated to obtain an affirmative identification will emit more behavioral cues than will administrators who are not motivated to obtain an affirmative identification.

Analysis Plan:

Individual Behavioral Cues: Administrators will receive a 0 or a 1 for each different behavioral cue. We will test the hypothesis using logistic regression analyses in which administrator motivation, lineup bias, and their interaction are included as predictors of each individual behavioral cue.
Total Behavioral Cues: We will sum the total number of behavioral cues encouraging the witness to choose emitted by each lineup administrator. We will also conduct a 2 (administrator motivation) x 2 (lineup bias) ANOVA on the mean number of behavioral cues emitted.

H1b: Suspect bias will moderate the extent to which the behaviors emitted by administrators will be directed toward the suspect.

Predicted Shape of Interaction: We expect to find a spreading interaction with a simple main effect of target of influence with more steering behaviors directed toward the suspect than fillers when suspect bias is high. We expect no significant difference of steering behaviors toward fillers or suspects when suspect bias is low.
Calculating Variables:
- We will sum the total number of behavioral cues emitted by each lineup administrator that are directed toward the suspect versus cues directed toward fillers.
- We will also calculate the proportion of behavioral cues emitted by each lineup administrator that are directed toward the suspect by dividing the suspect-specific cues by the total number of behavioral cues.

Data Prep Step 1: Processing Exclusions

First I processed exclusions from the data. Hover over the bars to see the remaining N after each step.

Our final sample size for coding analyses was N = 223.

Data Prep Step 2: Cleaning Variables

Next I completed some additional steps to ensure the data was clean and ready for analysis. This included

1. Converting specific behavior coding variables to numeric and replacing missing values with 0.
1. Recoding Bias_Condition (changing 1 to “F”) and converting both Bias_Condition and Motivation_Condition to factors.

Click the button below to view view the relevant r code.



``` r
  #Clean General Behavior Coding Data
  data <- data %>%
    mutate(across(c(Gen_Recall_YN, Gen_Quality_YN, Gen_MayDiff_YN, Gen_Familiar_YN, 
                    Gen_EncID_YN, Gen_DisID_YN, Gen_Careful_YN), 
                  ~as.numeric(as.character(.))))%>%
    mutate(across(c(Gen_Recall_YN, Gen_Quality_YN, Gen_MayDiff_YN, Gen_Familiar_YN, 
                    Gen_EncID_YN, Gen_DisID_YN, Gen_Careful_YN), 
                  ~replace_na(., 0)))

  #Clean Independent Variables
  data <- data %>%
    mutate(Bias_Condition = ifelse(Bias_Condition == 1, "F", Bias_Condition))
  data$Motivation_Condition <- as.factor(data$Motivation_Condition)
  data$Bias_Condition <- as.factor(data$Bias_Condition)
```

Overview of General Behaviors

Below is a list of our general (non photo-specific behaviors).

Behavior ID	Description
General Behaviors Reference
Gen_Recall_YN	Prompts recollection of culprit (e.g., "What do you remember about the culprit?", "Did he have a distinctive feature?", "Imagine him from another angle", "Describe/think about the features of the culprit", or similar)
Gen_Quality_YN	Asking about the quality of witness's view or memory (e.g., "Was the video grainy?", "Was it dark?", "Are you having trouble remembering?", "Can you remember clearly?", or similar)
Gen_MayDiff_YN	“The culprit might look different” (age, hairstyle, facial hair, clothes, etc.)
Gen_Familiar_YN	Prompts witness to consider the lineup members (e.g., "Does anyone/someone look familiar/stand out" or similar)
Gen_EncID_YN	Implying or stating the witness needs to make an ID (e.g., "You can pick any of these six"; witness asks if they have to pick someone & admin says "I think so"; "you need to pick someone" ; "pick whoever you think looks the closest"; "You can only pick one person" or similar) (DOES NOT include prompting questions such as "Does anyone look familiar?")
Gen_DisID_YN	Implying or stating the witness doesn't need to make an ID (e.g., "You don't need to pick anyone", “The culprit may not be present”, witness asks if they have to pick someone & admin says "No", or similar)
Gen_Careful_YN	“Take a good look” / “Look carefully” , “Take your time” / “No rush”, or similar

These were each dichotomous 0 or 1 questions to indicate whether the behavior did or did not occur at least once.

Overview of Photo Specific Behaviors

Below is a list of our photo-specific behaviors.

Behavior ID	Description
Photo-Specific Behaviors Reference
Ps_Attention	Asking about or drawing witness’s attention to a photo or photos (e.g., for a specific photo(s): sliding photo over/closer, tapping photo, pointing to photo, touching photo (but not to remove it), “you seem drawn to that photo(s)” or similar) (e.g., for general/no particular photo: “Do you want to take another look at the photo(s)?”, “Take another look”, or similar)
Ps_BringBack	Bringing photo(s) back into play after they were removed
Ps_Doubt	Expressing doubt in/questioning the witness’s preference/decision (e.g., Repeating the witness’s decision/preference in a questioning/skeptical way; “Are you sure/certain?”, “Is that your final choice/decision?”, “So you think it was [him/none of them]?”, “You’re removing these two?” or similar)
Ps_ReactPos	Reacting positively in response to a witness’s inclination toward identifying someone (e.g., smiling or indicating approval verbally or nonverbally).
Ps_Remove	Taking photos out of play (e.g., flipping over/removing photo(s))
Ps_Why	Prompts an explanation of the witness’s preference/decision (e.g., “Why do you think it’s him?”, “What features do you recognize?”, “Why do you think it’s none of them?” or similar)

For each photo-specific behavior logged, coders also indicated whether the behavior was directed toward/in response to…

A: the suspect
B: a filler
C: general
D: a lineup rejection
E: could not tell

General Behaviors : Individual Logistic Regression Analyses

Below we conduct a logistic regression analysis predicting the occurence each of the general behaviors. Click on the dropdown buttons from any of the analyses below to view the code or results.

1. Gen_Recall

Neither motivation nor lineup bias significantly predicted administrators’ Gen_Recall behavior.

Model Code

Gen_Recall_Mod <- glm(Gen_Recall_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-0.72	0.29	-2.5073657	0.012	-1.31	-0.17	0.49	6.29	1
Bias_ConditionF	-0.12	0.41	-0.2839098	0.776	-0.94	0.70	0.89	0.08	1
Motivation_ConditionControl	-0.22	0.41	-0.5354699	0.592	-1.04	0.59	0.80	0.29	1
Bias_ConditionF:Motivation_ConditionControl	-0.28	0.60	-0.4722838	0.637	-1.48	0.90	0.75	0.22	1

2. Gen_Quality

Neither motivation nor lineup bias significantly predicted administrators’ Gen_Quality behavior.

Model Code

Gen_Quality_Mod <- glm(Gen_Quality_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-2.55	0.52	-4.9024387	0.000	-3.74	-1.65	0.08	24.03	1
Bias_ConditionF	-1.41	1.14	-1.2383083	0.216	-4.40	0.55	0.25	1.53	1
Motivation_ConditionControl	-0.77	0.89	-0.8660167	0.386	-2.77	0.91	0.46	0.75	1
Bias_ConditionF:Motivation_ConditionControl	1.81	1.47	1.2328218	0.218	-0.88	5.24	6.12	1.52	1

3. Gen_MayDiff

Neither motivation nor lineup bias significantly predicted administrators’ Gen_MayDiff behavior.

Model Code

Gen_MayDiff_Mod <- glm(Gen_MayDiff_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-1.63	0.36	-4.475935	0.000	-2.41	-0.97	0.20	20.03	1
Bias_ConditionF	-0.63	0.59	-1.059917	0.289	-1.87	0.51	0.53	1.12	1
Motivation_ConditionControl	-1.26	0.70	-1.808353	0.071	-2.81	0.02	0.28	3.27	1
Bias_ConditionF:Motivation_ConditionControl	1.53	0.93	1.647398	0.099	-0.23	3.48	4.64	2.71	1

4. Gen_Familiar

Neither motivation nor lineup bias significantly predicted administrators’ Gen_Familiar behavior.

Model Code

Gen_Familiar_Mod <- glm(Gen_Familiar_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-0.56	0.28	-1.9964464	0.046	-1.13	-0.02	0.57	3.99	1
Bias_ConditionF	0.37	0.39	0.9415987	0.346	-0.40	1.15	1.45	0.89	1
Motivation_ConditionControl	0.10	0.39	0.2439865	0.807	-0.67	0.87	1.10	0.06	1
Bias_ConditionF:Motivation_ConditionControl	-0.55	0.55	-0.9920289	0.321	-1.64	0.53	0.58	0.98	1

5. Gen_EncID

Neither motivation nor lineup bias significantly predicted administrators’ Gen_EncID behavior.

Model Code

Gen_EncID_Mod <- glm(Gen_EncID_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-0.89	0.30	-3.00106713	0.003	-1.50	-0.33	0.41	9.01	1
Bias_ConditionF	-0.34	0.44	-0.76303368	0.445	-1.22	0.53	0.71	0.58	1
Motivation_ConditionControl	0.04	0.41	0.08514111	0.932	-0.78	0.85	1.04	0.01	1
Bias_ConditionF:Motivation_ConditionControl	-0.05	0.62	-0.07856213	0.937	-1.26	1.16	0.95	0.01	1

6. Gen_DisID

Neither motivation nor lineup bias significantly predicted administrators’ Gen_DisID behavior.

Model Code

Gen_DisID_Mod <- glm(Gen_DisID_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-2.85	0.59	-4.8042613	0.000	-4.27	-1.85	0.06	23.08	1
Bias_ConditionF	-0.39	0.93	-0.4133702	0.679	-2.44	1.45	0.68	0.17	1
Motivation_ConditionControl	0.89	0.72	1.2348286	0.217	-0.45	2.46	2.43	1.52	1
Bias_ConditionF:Motivation_ConditionControl	0.37	1.09	0.3346835	0.738	-1.77	2.68	1.44	0.11	1

7. Gen_Careful

Neither motivation nor lineup bias significantly predicted administrators’ Gen_Careful behavior.

Model Code

Gen_Careful_Mod <- glm(Gen_Careful_YN ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	0.18	0.27	0.6732670	0.501	-0.35	0.72	1.20	0.45	1
Bias_ConditionF	0.32	0.39	0.8124550	0.417	-0.45	1.09	1.37	0.66	1
Motivation_ConditionControl	0.28	0.38	0.7345744	0.463	-0.47	1.04	1.33	0.54	1
Bias_ConditionF:Motivation_ConditionControl	-0.92	0.54	-1.6899348	0.091	-2.00	0.14	0.40	2.86	1

Photo-Specific Behaviors : Individual Logistic Regression Analyses

First I created indicator variables to determine whether each photo-specific behavior occurred, regardless of what it was directed toward/in response to.

View the Relevant R Code

 data <- data %>%
          mutate(Ps_Attn_Photos_Occurred = if_else(
            coalesce(Ps_Attn_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_Attn_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_Attn_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_Attn_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_Attn_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))
        
        data <- data %>%
          mutate(Ps_BringBack_Photos_Occurred = if_else(
            coalesce(Ps_BringBack_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_BringBack_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_BringBack_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_BringBack_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_BringBack_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))
        
        data <- data %>%
          mutate(Ps_Doubt_Photos_Occurred = if_else(
            coalesce(Ps_Doubt_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_Doubt_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_Doubt_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_Doubt_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_Doubt_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))
        
        data <- data %>%
          mutate(Ps_ReactPos_Photos_Occurred = if_else(
            coalesce(Ps_ReactPos_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_ReactPos_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_ReactPos_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_ReactPos_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_ReactPos_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))
        
        data <- data %>%
          mutate(Ps_Remove_Photos_Occurred = if_else(
            coalesce(Ps_Remove_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_Remove_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_Remove_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_Remove_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_Remove_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))
        

        data <- data %>%
          mutate(Ps_Why_Photos_Occurred = if_else(
            coalesce(Ps_Why_Photos_1 == "Suspect", FALSE) | 
              coalesce(Ps_Why_Photos_2 == "Filler", FALSE) | 
              coalesce(Ps_Why_Photos_3 == "General/No Particular Photo", FALSE) | 
              coalesce(Ps_Why_Photos_4 == "Lineup Rejection", FALSE) | 
              coalesce(Ps_Why_Photos_5 == "I could not see which photo", FALSE), 
            1, 0
          ))

Then I conducted logistic regression analyses for each photo-specific behavior (whether it generally occurred or not, regardless of what it was directed toward/in response to).

1. Ps_Attn_Occurred

Neither motivation nor lineup bias significantly predicted administrators’ Ps_Attn_Occurred behavior.

Model Code

Ps_Attn_Occurred_Mod <- glm(Ps_Attn_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-0.48	0.28	-1.7360766	0.083	-1.04	0.05	0.62	3.01	1
Bias_ConditionF	0.67	0.39	1.7146401	0.086	-0.09	1.45	1.96	2.94	1
Motivation_ConditionControl	-0.06	0.39	-0.1464024	0.884	-0.83	0.71	0.94	0.02	1
Bias_ConditionF:Motivation_ConditionControl	-0.77	0.55	-1.4016264	0.161	-1.86	0.31	0.46	1.96	1

2. Ps_BringBack_Occurred

Neither motivation nor lineup bias significantly predicted administrators’ Ps_BringBack_Occurred behavior.

Model Code

Ps_BringBack_Occurred_Mod <- glm(Ps_BringBack_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-2.10	0.43	-4.855391	0.000	-3.06	-1.33	0.12	23.57	1
Bias_ConditionF	-1.85	1.10	-1.685599	0.092	-4.81	-0.04	0.16	2.84	1
Motivation_ConditionControl	-1.93	1.10	-1.754163	0.079	-4.88	-0.11	0.15	3.08	1
Bias_ConditionF:Motivation_ConditionControl	2.54	1.66	1.536657	0.124	-0.48	6.41	12.73	2.36	1

3. Ps_Doubt_Occurred

Neither motivation nor lineup bias significantly predicted administrators’ Ps_Doubt_Occurred behavior.

Model Code

Ps_Doubt_Occurred_Mod <- glm(Ps_Doubt_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	0.26	0.27	0.9413086	0.347	-0.27	0.80	1.29	0.89	1
Bias_ConditionF	0.09	0.39	0.2234462	0.823	-0.68	0.85	1.09	0.05	1
Motivation_ConditionControl	-0.65	0.38	-1.6880486	0.091	-1.41	0.10	0.52	2.85	1
Bias_ConditionF:Motivation_ConditionControl	-0.94	0.57	-1.6484822	0.099	-2.07	0.17	0.39	2.72	1

4. Ps_ReactPos_Occurred

Neither motivation nor lineup bias significantly predicted administrators’ Ps_ReactPos_Occurred behavior.

Model Code

Ps_ReactPos_Occurred_Mod <- glm(Ps_ReactPos_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Full Logistic Regression Table

term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-1.77	0.38	-4.62976643	0.000	-2.60	-1.08	0.17	21.43	1
Bias_ConditionF	0.31	0.52	0.60114029	0.548	-0.70	1.36	1.37	0.36	1
Motivation_ConditionControl	-0.04	0.54	-0.07716046	0.938	-1.12	1.03	0.96	0.01	1
Bias_ConditionF:Motivation_ConditionControl	-0.66	0.78	-0.85044050	0.395	-2.21	0.85	0.52	0.72	1

5. Ps_Remove_Occurred (Significant)

Motivation condition significantly predicted administrators’ Ps_Remove_Occurred behavior.

Model Code

Ps_Remove_Occurred_Mod <- glm(Ps_Remove_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Logistic Regression Table: Remove Photos
term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-1.39	0.34	-4.1124137	0.000	-2.10	-0.76	0.25	16.91	1
Bias_ConditionF	0.05	0.48	0.0973497	0.922	-0.90	0.99	1.05	0.01	1
Motivation_ConditionControl	-1.50	0.68	-2.2045849	0.027	-3.04	-0.27	0.22	4.86	1
Bias_ConditionF:Motivation_ConditionControl	1.01	0.85	1.1875266	0.235	-0.60	2.81	2.75	1.41	1

6. Ps_Why_Occurred (Significant)

Lineup bias significantly predicted administrators’ Ps_Why_Occurred behavior.

Model Code

Ps_Why_Occurred_Mod <- glm(Ps_Why_Photos_Occurred ~ Bias_Condition + Motivation_Condition + Bias_Condition*Motivation_Condition, family = binomial(), data = data)

Logistic Regression Table: Why Prompt
term	B	SE	statistic	p	95% CI Lower	95% CI Upper	OR	Wald_Chi_Sq	df
(Intercept)	-1.50	0.35	-4.3022448	0.000	-2.25	-0.86	0.22	18.51	1
Bias_ConditionF	1.54	0.44	3.4674233	0.001	0.70	2.45	4.67	12.02	1
Motivation_ConditionControl	-0.64	0.56	-1.1450817	0.252	-1.78	0.43	0.53	1.31	1
Bias_ConditionF:Motivation_ConditionControl	-0.64	0.70	-0.9257546	0.355	-2.00	0.75	0.53	0.86	1

ANOVA #1 — Total Number of Behaviors (Significant)

1. Prepping the Variables

First I created a variable summing the number of behaviors which occurred. Note, I incorporated only the general occurrence (yes/no) of each photo-specific behavior, regardless of what photo the behavior was in reference to. IMPORTANT NOTE:, This does not actually reflect the full number of instances of each behavior — that information is in our open-ended coder data. Think of this more as a reflection of the number of different types of behaviors which occurred. So larger numbers = larger breadth/variation in the different types of behaviors administrators used.

View Code for Computing Total Behaviors Variable

data <- data %>% mutate(
          NumTotalBehaviors = rowSums(across(c(Gen_Recall_YN, Gen_Quality_YN, Gen_MayDiff_YN, Gen_Familiar_YN, 
                                               Gen_EncID_YN, Gen_DisID_YN, Gen_Careful_YN, Ps_Attn_Photos_Occurred,
                                               Ps_BringBack_Photos_Occurred, Ps_Doubt_Photos_Occurred,
                                               Ps_ReactPos_Photos_Occurred, Ps_Remove_Photos_Occurred, 
                                               Ps_Why_Photos_Occurred))))

2. Descriptives

At a glance, it seems motivated administrators were utilizing a broader range of different behaviors than unmotivated administrators.

Total Behaviors By Motivation and Bias
Motivation_Condition	Bias_Condition	Mean	SD	Min	Max
$20 motivation	B	3.400000	2.232753	0	9
$20 motivation	F	3.811321	2.029178	0	8
Control	B	2.877193	1.852317	0	8
Control	F	2.603448	2.127016	0	8

Total Behaviors By Motivation
Motivation_Condition	Mean	SD	Min	Max
$20 motivation	3.601852	2.135327	0	9
Control	2.739130	1.991592	0	8

Total Behaviors By Bias
Bias_Condition	Mean	SD	Min	Max
B	3.133929	2.055516	0	9
F	3.180180	2.158275	0	8

3. ANOVA

Next I ran the ANOVA. There was a significant effect of motivation condition on number of behaviors employed. Motivated administrators employed a larger number of different behaviors than unmotivated administrators.

View Code for the ANOVA Model

        TotalBehaviorsMod <- anova_test(NumTotalBehaviors ~ Bias_Condition * Motivation_Condition, data = data, type = '3', effect.size = "ges")

Total Behaviors ANOVA Table
Effect	DFn	DFd	F	p	p<.05	ges
Bias_Condition	1	219	0.062	0.804		0.000282
Motivation_Condition	1	219	9.784	0.002	*	0.043000
Bias_Condition:Motivation_Condition	1	219	1.533	0.217		0.007000

ANOVA #2 — Proportion of Suspect Behaviors (Significant)

1. Prepping the Variables

First, I created a variable, NumSuspectBehaviors, that sums the total number of behaviors specifically directed at the suspect across all photo-specific administrator behaviors.

View Code for Computing Total Suspect Behavior Variable

        data <- data %>% mutate(
          NumSuspectBehaviors = rowSums(across(c(Ps_Attn_Photos_Suspect, Ps_BringBack_Photos_Suspect,
                                                 Ps_Doubt_Photos_Suspect, Ps_ReactPos_Photos_Suspect,
                                                 Ps_Remove_Photos_Suspect, Ps_Why_Photos_Suspect))))

Next, I created a proportion variable, PropSuspectBehaviors, which captures the proportion of total behaviors that were directed at the suspect out of the total number of photo-specific behaviors exhibited.
NOTE: This proportion will not be inversely related to proportion of filler-directed behaviors, because behaviors could also have also been neither filler nor suspect directed, but instead might have been general (no particular photo), in response to a lineup rejection, or “could not tell which photo”.
NOTE: If no photo-specific behaviors were exhibited during the lineup, I assigned a value of 0 to the suspect behavior proportion.

View Code for Computing Proportion Suspect Behavior Variable

        data <- data %>% mutate(
          PropSuspectBehaviors = ifelse(NumTotalBehaviors > 0,
                                        NumSuspectBehaviors / NumTotalBehaviors, 
                                        0)
        )

2. Descriptives

Below are grouped descriptives for proportion of suspect-directed behaviors. At first glance, it seems there are higher proportions of suspect-directed behaviors when administrators are motivated.

Proportion Suspect Behaviors By Motivation and Bias
Motivation_Condition	Bias_Condition	Mean	SD	Min	Max
$20 motivation	B	0.1951371	0.2930464	0	1.0
$20 motivation	F	0.2000898	0.2721181	0	1.0
Control	B	0.1349833	0.2287852	0	1.0
Control	F	0.1137726	0.1806566	0	0.5

Proportion Suspect Behaviors By Motivation
Motivation_Condition	Mean	SD	Min	Max
$20 motivation	0.1975676	0.2816587	0	1
Control	0.1242857	0.2052901	0	1

Proportion Suspect Behaviors By Bias
Bias_Condition	Mean	SD	Min	Max
B	0.1645231	0.2628634	0	1
F	0.1549871	0.2319315	0	1

3. ANOVA

Next I ran the ANOVA. There was a significant effect of motivation but not lineup bias on proportion of suspect-directed behaviors employed.

View Code for the ANOVA Model

        SuspectBehaviorsMod <- anova_test(PropSuspectBehaviors ~ Bias_Condition * Motivation_Condition, data = data, type = '3', effect.size = "ges")

Proportion Suspect Behaviors ANOVA Table
Effect	DFn	DFd	F	p	p<.05	ges
Bias_Condition	1	219	0.061	0.806		0.000277
Motivation_Condition	1	219	4.925	0.027	*	0.022000
Bias_Condition:Motivation_Condition	1	219	0.157	0.692		0.000717

Aronson APLS Analyses

Ellie Aronson

2024-09-22

Hypotheses Overview

Data Prep Step 1: Processing Exclusions

Data Prep Step 2: Cleaning Variables

Overview of General Behaviors

Overview of Photo Specific Behaviors

General Behaviors : Individual Logistic Regression Analyses

1. Gen_Recall

2. Gen_Quality

3. Gen_MayDiff

4. Gen_Familiar

5. Gen_EncID

6. Gen_DisID

7. Gen_Careful

Photo-Specific Behaviors : Individual Logistic Regression Analyses

1. Ps_Attn_Occurred

2. Ps_BringBack_Occurred

3. Ps_Doubt_Occurred

4. Ps_ReactPos_Occurred

5. Ps_Remove_Occurred (Significant)

6. Ps_Why_Occurred (Significant)

ANOVA #1 — Total Number of Behaviors (Significant)

1. Prepping the Variables

2. Descriptives

3. ANOVA

ANOVA #2 — Proportion of Suspect Behaviors (Significant)

1. Prepping the Variables

2. Descriptives

3. ANOVA

Data Visualizations