Is it a factor structured scale or have you generated it yourself? If you generated it yourself, you could run an Exploratory Factor analysis and try to create a 6-item scale from there, and then run a Confirmatory test to confirm the factor structure.

Is this an existing scale or a new scale I suppose is the first question.

Is it a factor structured scale or have you generated it yourself? If you generated it yourself, you could run an Exploratory Factor analysis and try to create a 6-item scale from there, and then run a Confirmatory test to confirm the factor structure.

Is this an existing scale or a new scale I suppose is the first question.

I agree with Michael. Technically, run a confirmatory factor analysis (CFA) and deselect step-by-step one item after another item (e.g. by the criteria of the lowest factor loading or something similar) as long as you achieve the 6 items solution but only if the model fits the data. By my question is: why do you want to have exactly 6 items?

Given your assumption of a single-factor model, I would run a Cronbach's alpha before an Exploratory Factor Analysis (in SPSS it is called Reliabilities). Be sure to ask for the "alpha if item deleted" option, and then drop the items one by one until you start getting serious decreases in the reliability. (The option for item-to-total correlation will also give you similar information.)

But I agree with Gert's question -- why is six your target for the number of items?

Dear All,

Thank you very much for your kind answers and solutions.

I've made a questionnaire which is contain 34 criteria and all the respondents answer the questions.

I've done this because I need to reduce the number of criteria to maximum of 10(According to the outcome of the questionnaire). In the next step, I'm going to use AHP method to make another questionnaire to do pairwise analysis. (The maximum of 10 criteria can utilize for AHP method)

At the end of my study, I need to find the weight of each criteria. Am I in a right way??

Kindest regards,

Reza

1-You can use prearson correlation between each item and the total scores

2-Sort the total scores Ascending college grades (or descending) then select 27% of the sample size have the minimum scores and 27% of the sample size have the maximum scores then, apply t-test for two independent samples. The six items which have the largest values of t-test represented the 6 most important items.

3- You can apply the factor analysis.

A simple mean score (MS)analysis , relative importance index (RII) or severity index (SI)analysis can be used to rank the factors. if the scale is from 1-lowest to 9-highest, then the factor with the highest MS, RII or SI is ranked 1 and the factor with the least is ranked 21. if the scale is from 1-highest to 9-lowest, then the factor with the lowest MS, RII of SI is ranked 1 and the factor with thehighest is ranked 21.In this way you pick the factoes ranked fro 1 to 6. MS, RII and SI will give the same results, i.e pick the same 6 factors. These methods are used extensively and are common in the literaure

سلام اگر می تونید فارسی توضیح بدید جملات مفهوم نیست.

اگر درست فهمیده باشم راه حل

Adebayo Akanbi Oladapo

درست است.

Thank for all constructive comments. I have a set of variables and just intend to keep most important ones and drop the rest. The nature of the data does not allow factor analysis as the level of variance is low. I can use RII but i like to use a more robust method.

Huda: Would you please guide me towards a reference?

Adebayo: I have used RII in many occasions but SI is usually used in medical research. Are you aware of any research in the construction context?

Kind Regards,

Reza

The Relative Importance Index, is the best way to go. I like how it automatically ranks the variables under concern. But cut off points determination is one of the challenges with the RII.