Subsample Selection Bias :
Dear all;
I have a sample of 3000 paintings, then I divided this sample into 5 subsamples given some criteria ( year, style ...). Then
I did some ANOVA for each subsample in order to analyse consumer preferences.
The Reviewer 1 comment is beleow:
How can the authors warrantee that the specific selection of paintings is not responsible for the effect?
It is not clear to me how selection bias is ruled out.
1-How can I answer the Reviewer here?
2- Should I use bootstrapped ANOVA to correct my results?
3- Is there anyway to handle Subsample Selection Bias in this case?
Thank you!
Hello Aymen,
Well, the good news is, you appear to have the opportunity to respond to a concern about the study, as opposed to having it rejected outright by the peer review process!
It's not completely clear to me just what you did, so I apologize if this isn't particularly helpful.
1. The usual methodological concern when people mention selection bias is that of a purposive arrangement of cases so that a given result is more likely to occur. Without understanding the execution of the study, it's hard for me to respond to this concern.
What was/were the independent variable(s), and what was/were the dependent variable(s)? Were all 3000 paintings included in the 5 subsamples (e.g., 600 per subsample)? Did participants rate one painting each, all paintings in a single subsample, or all 3000 paintings?
Were the subsamples formed randomly or in a stratified random manner from the paintings that met whatever selection criteria you applied?
2. For the second question, bootstrap/resampling/or exact tests won't overcome any bias; they just change the nature of the question by reframing the comparison to that of the universe of possible outcomes for the observed data points (usually under different assortments associated with the independent variable/s) instead of to some theoretical distribution with a presumed shape and location.
3. Unfortunately, can't answer this without knowing more about this study.
Good luck with your work!
for respondents you can check non responsive bias.
Refer the link:
https://www.tandfonline.com/doi/abs/10.1080/14783363.2016.1230470?journalCode=ctqm20
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA