IALSA / wave-inclusion

How does the number of waves included into the analysis affect the conclusions from a longitudinal study?
GNU General Public License v2.0
0 stars 0 forks source link

Q1. Wave Inclusion and Conclusions about Change #4

Open andkov opened 8 years ago

andkov commented 8 years ago

Using the existing reports from the MAP dataset kb profiles (one slate for each pair of processes) and kb_waves (many scenarios for a single pair), find the words to beging addressing the following question:

Would the conclusions drawn about cognitive change have been different had we analysed the MAP data at different points of the study's lifespan.

andkov commented 8 years ago

Is wave inclusion even an issue?

How do we demonstrate that is

casslbrown commented 8 years ago

Across 16 different cognitive performance tasks completed by Rush MAP participants there were numerous examples of linear trajectories that were not significant with fewer occasions of measurement that became significant once a greater number of occasions were included in the analysis and some examples of the reverse. Examining changes in the slope parameter estimate, indicating rate of linear change, for each cognitive outcome there were 4 ( category fluency, digit symbol, MMSE, and number comparison) out of 16 cognitive measures where the conclusions drawn about change over time would have been significantly different depending on how many assessment occasions were included in the analysis for both men and women. There are an additional 4 cognitive measures (BNT, complex ideational material, digits backwards, and digits forward) where interpretations would have changed only for women, and 2 cognitive measures (boston story delay and logical memory immediate recall) where the interpretation changed only for men with the addition of more assessment occasions. Thus, conclusions drawn about changes in cognitive performance over time for Rush MAP participants would be different for a number domains depending on how many waves of data collection were available at the time and/or how many were included in the analysis.