Applied Linguistics at KMUTT

FROM: Phuttharaksa Yantraprakorn (12/02/09 7:25 PM PST)
SUBJECT: Validity and Reliability

Can there be validity without reliability?

FROM: woravut j (12/02/09 9:55 PM PST)
SUBJECT: RE: Validity and Reliability

There is a very good summary of validity and reliability. It covers different views on validity and reliabiliy, for example,
- conventional views of reliability (AERA et al., 1985, )
- a modified view of reliability (Moss, 1994),
- an extended view of Moss's reliability (Mislevy, 2004),
- a radical view of reliability (Thompson et al, 2003),
- an updated perspective of reliability (Cronbach, 2004)

Just follow the link if you want to know more.

http://www.creative-wisdom.com/teaching/assessment/reliability.html

Woravut

FROM: Richard Watson Todd (12/03/09 5:37 PM PST)
SUBJECT: RE: Validity and Reliability

The webpage Woravut suggests is interesting and useful. But (there's always a 'but') it concerns the validity and reliability of tests. Are these the same as the validity and reliability of research? If not, how do they differ?

FROM: Thanakorn Thongprayoon (12/04/09 5:08 AM PST)
SUBJECT: RE: Validity and Reliability

With some of my readings, I have found that the real difference between reliability and validity is mainly a matter of defining them. Reliability estimates the consistency of the measurement, or more simply the degree to which an instrument measures the same way each time it is used in under the same conditions with the same subjects. Validity, OTOH, involves the degree to which you are measuring what you are supposed to, more simply, the accuracy of your measurement.

To answer the question asked. The research findings can be valid, but reliable. In case you use the most suitable research tools to measure what they are supposed to, but you use them in different aspects, at different times, with different subjects, and under different conditions. Then you might get different things to do after that.

They, reliability and validity, are like spoon and fork. If you want to eat as others do and accept, so you need to use them right together and carefully. Not only for better taste, but also looking good.

FROM: woravut j (12/07/09 5:32 PM PST)
SUBJECT: RE: Validity and Reliability

It seems you have a focus on the use of the instrument. But it may be possible that we look at validity and reliability in 4 different layers.

The instrument
The use of the instrument
The result from the use (of the instrument)
The interpretation of the result

FROM: Thitirut Wijanpreecha (12/02/09 7:42 PM PST)
SUBJECT: Naturalistic Approach

For example, as Ethnography studies human belief, culture or way of life.

Do researchers need to replicate (reliability)and collect data on large scale (validity)?

FROM: woravut j (12/02/09 10:22 PM PST)
SUBJECT: RE: Naturalistic Approach

To answer this, we may need to think about the principles of ethnography and how ethonographers conduct their research.
.......
It may take many months to a few years to conduct an ethnography study.

The world is changing very quickly, and it may be different to live in the same way in today's world as we used to in a previous year. Some progress (or regress?) occurs every second.
.....

woravut
PS. Replication is a must, but in a laboratory

FROM: Thitirut Wijanpreecha (12/02/09 7:59 PM PST)
SUBJECT: IR & ER

If researchers would like to determine replicability and consistency of obtained data, they have to conduct it to the same participants or could recheck it with other similar group.

FROM: woravut j (12/02/09 10:05 PM PST)
SUBJECT: RE: IR & ER

Replicability: It is difficult. Though we may have the same group of participants but their knowledge, skills, emotions, attitudes are not stable. So, it is also more difficult with other groups.

Woravut

FROM: Thanakorn Thongprayoon (12/04/09 5:20 AM PST)
SUBJECT: RE: IR & ER

Research instruments, treatments, and research questions are likely to be replicated with the same or similar process rather than the same subjects I guess. As the subjects may get affected from factors, having been avoided (not controlled) in the previous study.