Floor and Ceiling Effect

[Mehwish]

Hello MedStatisticians,

We have conducted a research on Post-traumatic stress on School teachers after a terrible terrorists' attack on school children in Pakistan on December 16, 2014. We have used 3 scales, two were standard WHO-Stress Scale, Davidson Trauma Scale (DTS) and a Self-Constructed Perception Scale on specific scenario. DTS has pre-defined 5 sub-scales.

We got data of 390 teachers and we are going to present in a national symposium. The responses on scales and sub-scales were somewhat skewed. I read somewhere that Floor and Ceiling Effect can be calculated.

These measures are new to me. I got many definition and description while online browsing. Though, I am unable to comprehend the calculation of these effects much clearly. I am copying the related extractions upon my search and I need your help for the same calculations for my scales.

1. Ceiling effect: proportion of patients scoring at the highest possible scale level. Floor effect: proportion of patients scoring at the lowest possible scale level. (http://intqhc.oxfordjournals.org/content/17/6/465.full#T3)

2. A ceiling effect occurs when a measure possesses a distinct upper limit for potential responses and a large concentration of participants score at or near this limit (the opposite of a floor effect). Scale attenuation is a methodological problem that occurs whenever variance is restricted in this manner. The problem is commonly discussed in the context of experimental research, but it could threaten the validity of any type of research in which an outcome measure demonstrates almost no variation at the upper end of its potential range. Ceiling effects can be caused by a variety of factors.(http://srmo.sagepub.com/view/the-sage-encyclopedia-of-social-science-research-methods/n102.xml). A floor effect occurs when a measure possesses a distinct lower limit for potential responses and a large concentration of participants score at or near this limit (the opposite of a ceiling effect). Scale attenuation is a methodological problem that occurs whenever variance is restricted in this manner. The problem is commonly discussed in the context of experimental research, but it could threaten the validity of any type of research in which an outcome measure demonstrates almost no variation at the lower end of its potential range. Floor effects can be caused by a variety of factors.(http://srmo.sagepub.com/view/the-sage-encyclopedia-of-social-science-research-methods/n344.xml) 

3. High floor and ceiling effects indicate insensitivity for detecting a change of symptoms and the maximum cut-off for floor and/or ceiling effects should be no more than 15% [24] (http://www.hindawi.com/journals/ijr/2013/506518/).

4. An answer from ResearchGate on calculation: Proportion of patients on minimum and maximum values of scale is floor and ceiling effect respectively.

So, my questions are should I show the proportion of teachers at minimum and maximum score of my 3 scales and 5 subscales? Or should I mention the score of scale at, say at-most 10% and at-least 90%, of proportion of teachers? What other possible answers can floor and ceiling effects can provide?

Thank you in advanced,

[Adrian Sayers]

It depends on what you believe, some people say a true ceiling is 100% some say 95%.

To air on the side of transparency i would probably illustrate the proportion at 0% & 10% and 90% & 100%.

However, if you have mild floor ceiling effects you can correct them with Tobit / mixed distribution based models.

GLAMM is pretty good at doing this.

bw
Adrian

--
--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules

---
You received this message because you are subscribed to the Google Groups "MedStats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to medstats+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Mehwish Hussain]

Thank you Adrian, can you please tell me the suggested percentages were on scale score or on proportion of participants?

--

Regards,

Mehwish Hussain, PhD*
Senior Lecturer of Biostatistics
Manager, Research Development

Office of Research, Innovation and Commercialization

Dow University of Health Sciences

Karachi, Pakistan

G

o

o

g

l

e

 

Scholar

:

 

http://scholar.google.com.pk/citations?user=o5Nz6PwAAAAJ&hl=en&oi=ao

​

Research Gate: https://www.researchgate.net/profile/Mehwish_Hussain/

​

​

LinkedIn: http://www.linkedin.com/pub/mehwish-hussain/17/a8a/449

​

[Adrian Sayers]

Percentage Of the score.

Bw 

Adrian

[Mehwish Hussain]

OK Thanks, got it.