'Differences in implicit learning between cognitively impaired, depressive and healthy elderly using the Serial Reaction Time Task (SRTT)' (AsPredicted #84388)
Author(s) This pre-registration is currently anonymous to enable blind peer-review. It has 5 authors.
Pre-registered on 01/05/2022 09:32 AM (PT)
1) Have any data been collected for this study already? No, no data have been collected for this study yet.
2) What's the main question being asked or hypothesis being tested in this study? Is there a difference in implicit learning of individuals who are not cognitively impaired in comparison to subjectsindividuals with dementia, and individuals with depression?
To address this question, differences in (i) average response timereaction times, (ii) accuracy, and (iii) implicit learning (i.e., reaction time difference between sequence and random blockssee dependent variables) between the three groups will be analyzed.
Hypothesis
(i) We predict that cognitively impaired subjects are on overage slower than healthy controls.
(ii) We predict that cognitively impaired subjects conduct more errors than healthy controls.
(iii) We predict that the implicit learning is less pronounced in cognitively impaired subjects compared to healthy controls.
Differences in implicit learning, reaction time, and accuracy between individuals with depressive symptoms and healthy as well as cognitively impaired individuals will be explored.
3) Describe the key dependent variable(s) specifying how they will be measured. • Reaction time and accuracy (correct/incorrect) per trial
• Median reaction time: Median reaction time per subject per block
• Average reaction time: The average reaction time across the experiment per subject, computed as the average of the median reaction time of each block.
• Accuracy: The proportion of correct trials to the total number of trials per subject per block
• Implicit learning: the response increase between the estimated reaction time, based on the estimated learning curve during the training phase, and the measured reaction time in the random block, coded as following:
response_increase=1 if random block, else 0, i.e,
block response_increase
0 0
1 0
2 0
3 0
4 1
4) How many and which conditions will participants be assigned to? Overall: 90, 30 per group
• Healthy Control (HC): not cognitively impaired / no dementia / no depression
• MCI/Dementia: cognitively impaired or dementia / no depression
• Depression: not cognitively impaired / no dementia / depression
Participants must be at least 60 years old and have not participated in preceding studies using a similar method.
5) Specify exactly which analyses you will conduct to examine the main question/hypothesis. General modelling approach (ii + iii): Mixed effect models will be hierarchically constructed from a full model (containing all fixed and random effects including interactions) with a maximum random effects structure to the model with the best fit according to Bayesian Information Criterion (BIC) by removing the most complex fixed effects first (i.e., interaction terms). If the complexity of the random effect structure is not supported by the data (i.e., convergence issues), the random effects structure will be reduced similar to the fixed effects structure, by removing the most complex terms first. Coefficients and their significances will only be extracted for the best fitting model according to the outlined hierarchical approach after the best fitting model has identified found.
(i) Average reaction time: Analysis of Covariance (ANCOVA) will be performed on the cleaned data to contrast the difference on the average response time across the experiment. Fixed effects will be group, with age, gender, education years as covariates.
Based on recommendations for good scientific practice for reporting ANCOVAS (Simons et al., 2011) an ANOVA comparing the average reaction times between groups will be performed (showcasing the impact of the covariates on the results). In case of non-normal distribution of the residuals, non-parametric tests will be used instead of ANCOVAs.
(ii) Accuracy: Generalized Linear Mixed Effects Models (GLMEs) will be run on the raw data to contrast the difference in accuracy between the groups. Fixed effects will be group, block, and their interaction. Random effect will be participant.
(iii) Learning curve and implicit learning: LME spline models with the last training block as the knot will be performed on subject's median reaction times to contrast the influence of group, block (learning curve), and response increase (implicit learning). Fixed effects will be block, response increase, group, gender, age, and education, and the interaction between block and group and response increase and group. Random effects will be participant. Linear, quadratic, and cubic effects of block will also be tested for their contribution to model fit.
If the residuals of the LMER models are non-normally distributed, we will use z-transformation on the raw reaction times.
Classification: A random forest will be used to predict the subject's group (i.e., cognitively impaired vs. depressive vs. healthy). Input features will be variables from the analyses outlined above.
In an exploratory analysis, we will run Bayes factor and LME analyses on a combined data set of the collected data and data from previously conducted studies.
6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations. We exclude subjects
- who fail at the test runs preceding the main experiment three times
- whose accuracy is below 70% across the experiment
- whose data is incomplete
- with discontinuation during assessment
Data cleaning: for reaction time analyses, we will remove trials within blocks as follows:
• First trial of each block
• Trials with reaction times below 200ms
• Erroneous trials
• Trials following an erroneous trial
• Trials with reaction times deviating more than 2.5 standard deviations from the mean within a block (within subjects)
7) How many observations will be collected or what will determine sample size? No need to justify decision, but be precise about exactly how the number will be determined. We aim at 90 subjects in total (30 subjects per group). The final sample size depends on the available number of cognitively impaired subjects/subjects with dementia and matched healthy controls and subjects with depression.
Sample size rationale: Especially due to the personnel limitation for data collection and the availability of the clinical sample a larger number cannot be recruited.
Stopping rule: 30 participants per condition (complete data).
8) Anything else you would like to pre-register? (e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?) Besides the Serial Reaction Time Task (SRTT), we will also collect data with the Montreal Cognitive Assessment Test (MoCA) and Geriatric Depression Scale (GDS).