Referees usually asks about the existence of ceiling effect or floor effect in the process of instrument development. I am interested to find the way I can statistically assess them.
the ceiling and flooring effects were calculated by percentage frequency of lowest or highest possible score achieved by respondents. the ceiling and flooring effects of more than 15 % were considered to be sig.
I agree with Alvin, you basically want to determine that is sufficient variance in your observations. Variables that are strongly skewed basically indicate that everyone is answering the question the same way, using one end or the other of the scale.
I don't think that there is a statistical method suited to detect ceiling and floor effects. In addition to the suggestion you already received, I would count the responses to the lowest and the highest possible value and report the percentage.
In case you encounter these effects, you would need to use methods that take the censoring into account, such as Tobit regression, OLS regression, or truncated regression. These methods all have advantages and disadvantages and cannot be used for all types of censored or truncated data.
HTH, Karin
Thanks all. But the important question is that what percentage is acceptable?
for example if 30% of respondents chose 1 for an item with 4 choice (1-4), may I report that there is a floor effect?
Well, yes. Report skewness and kurtosis as well as the percentage of extreme values. The percentage is continuous, therefore cut-off-values do not exist.
Regards, Karin
the ceiling and flooring effects were calculated by percentage frequency of lowest or highest possible score achieved by respondents. the ceiling and flooring effects of more than 15 % were considered to be sig.
Hi, do you have any reference for cut-off for ceiling and flooring effects? Thank you in advance.
Bojana Dinić
In this article there's some information regarding cut-offs, as well as reference for another article.
Article Quality criteria were proposed for measurement properties of...
With some of the quality of life questionnaires in chronic diseases, we often find this issue. How to address this?
I am dealing with similar questions. What I had found is that you can run some interesting codes in R.
rec.mean.var(dataframe$variable) #estimates true mean and variance of the data with ceiling/floor effects
dataframe$variableceiling
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire