To my understanding, instrument development implied used of relevant variables to construct statements (items) based on the purpose for which the instrument is been constructed.

Once that is done, the validity and reliability can be determined following the bellowed basic steps:

1. Face and content validity.

2. Construct validity

- item reductions ( EFA)

- validity testing

3. Reliability testing

- internal consistency

(Cronbach's alpha)

- stability of the instrument

overtime.

It should also be noted that either the CTT or IRT approach might be used during steps 2 & 3.

Hope thislittle contribution might be of help. Thanks

Hello

As far as i understand scale development, the following steps can be taken

1. Operational definition of the construct

2. Are there other scales available ? If yes... whats the need to develop a new one.

3. Theory Consultation / Review of Literature

4. Item Writing

5. EFA

If you want to validate the tool. you need to perform CFA with the help of any tool available on a different sample.

Last step would be Convergent and Divergent Validity of the newly developed tool.

(Cronbach's Alpha would be an add on)

Hi Soumaya Elhassouni

Here is your questions and my answers:

1/When we say "developing" (not validating), does it mean that using one set of data would be enough?

Multiple data sets may be needed to improve the scale. The sample you use to develop a scale is important. The sample should be selected to provide maximum variance according to the purpose of the scale.

3/Exploratory=development? Vs confirmatory =validating ??

EFA and CFA uses for both scale development and validating. The EFA is used more frequently to examine the relationships between the items and to make the operational definition of the intended measure. Also EFA and CFA are used for validation process. There are related to construct validity. Both EFA and CFA are also used in the validation and scale development process.

4/ What are the statistical methods usually used in the developing step (again not validation): EFA with factorisation tests alone (Bartlett, KMO...)?

There are many methods. But EFA and CFA must be used to understand latent variable. In developing step, EFA is used to understand latent variable. In other words, EFA helps us to understand which item is related to which dimension. Then dimensions are named. In addition to this, the opinions of the subject experts about the items are taken during the scale development process. Item discrimination is examined. It is examined whether the items are understood by the group. The EFA is part of the scale development process.

5/Is Cronbach's alpha used in the exploratory step? Confirmatory step? Both of them?

Reliability can be used all steps. But it is important to control final scale Cronbach alfa (or other strong reliability indexes) value. It is not restricted in a step.

6/Is Amos a compulsory step in validating or we could just use another factor analysis in a second set of data with additional statistical tests (e g Joreskog?)

AMOS is a software only. It would be better if the following development process would be followed:

1) EFA is done by collecting data from the sample. The scale is developed and finalized with other processes (subject expert opinions vs)

2) Data is collected from a different sample with the final scale and CFA is performed.

7/Is developing a scale what we call "measurement model"? Which means no independant, moderating or mediating variables are involved, right?

Measurement model is established via scale develeopment process. When other variables are added, it is named structural model.

Thank you very much all of you for your precious answers, I will take time to understand your explanation and maybe will bother you again! 1000 of thanks!

Try this App

https://play.google.com/store/apps/details?id=inferentialstatistics.myapplication