Re: Scale or Items?

[Art Kendall-2]

Is this a single scale or are there 2 scales with different reliabilities? or is that pre and post?

Since the scales have poor reliability, do you have specific items you are interested in (i.e., have hypotheses for)?  Is there a subset of the items which does sum into a scale with better reliability?

Are there other variables you want to use in your model? Did you have a specific intervention, or is your interest strictly developmental?


How many cases do you have?

Can you match the pre and post measures within each case?
How much attrition was there between the two measures?

What is the response scale on the items?

It is possible that you can use a more elaborate kind of general linear model than the t-test, and use modified t-tested for post hoc comparisons, but a lot depends on the answers to the questions above.

Art Kendall
Social Research Consultants

Samaa Taher M. Attia wrote:

 

Dear Prof. Kendall,

Thank you so much for your kind help. Yes, the reliability is OK 0.6 and 0.7, but what I am asking is the data to address, can I use the items?

 

In more details: I am comparing the communication apprehension between first and final year, so I am using t test. But here, I want to use the communication apprehension items/data rather than the scale (as you know, we have oral and written). Can I use the items rather than the sacale? knowing that I did not find any study used the items before, all of them used scales.

 

Really, thank you so much and waiting for your kind reply,

Samaa

 

--- On Sun, 7/26/09, Art Kendall [hidden email] wrote:
,

From: Art Kendall [hidden email]
Subject: Re: Scale or Items?
To: "Stat-Help :)" [hidden email]
Date: Sunday, July 26, 2009, 3:04 PM

At a minimum you should use RELIABILITY to check that the syntax you are using to create the summative score implements the scoring key correctly.

What questions are you asking the data to address?


Art Kendall
Social Research Consultants

Stat-Help :) wrote:
> Hello all,
>
> This is my first contribution to this group, hope you can help me.
> I am doing a study on "Communication Apprehension", however, when I
> searched all previous studies, I found them used it as scales, but I
> would like to use it as items as well as scales. Is this possible, am
> I right?
>
> Your help is highly appreciated and thanks :))
>
> S.A

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[Art Kendall-2]

so far it sounds like you have 2 independent between group variables.� Year: First year vs fourth year students, Major: marketing vs business&admin.
Is that correct?
Your response scale was 1 to 4 for some items and 1 to 5 for others.� Or it was 1 to 5 for all items but for some items nobody answered 5.
What were the verbal anchors on your response scale?�


Are the constructs for written vs oral the same? What are the constructs the instrument was designed to measure?
Did the originators of the scale factor analyze the scale?

What are the reliabilities of the subscales?

How many separate dependent variables were used in the previous research? 1 grand total?� 2 oral and written? 8?

Art Kendall
Social Research Consultants

Samaa Taher M. Attia wrote:

Is this a single scale or are there 2 scales with different reliabilities? Oral communication apprehension has four subscales. And Written communication apprehension has also four subscales.

or is that pre and post? because pre and post needs much time, I followed this approach: � I am collecting data from first and final year students as if first year students are “ Pre” while final year students are “ Post”

Since the scales have poor reliability, do you have specific items you are interested in (i.e., have hypotheses for)?� That is very interesting question, in the previous literature, they used scales, none of them used the items, however, I wanted to get a much closer look, so I wanted to use items. However, I do not have hypotheses for specific items, no, instead, as you know, few items form a certain scale, so, I mentioned those items in specific (all of them) not certain items.

Is there a subset of the items which does sum into a scale with better reliability? I tried, but based on the sample I have, no. instead, I justified it by saying: 1) it is relatively new use for the scale in this context, and therefore, it could be looked at as exploratory study

Are there other variables you want to use in your model? Yes, but it is for different hypothesis. The second one. For this second one, I am using the scale not the items.

Did you have a specific intervention? No.

Or is your interest strictly developmental? Yes, developmental


How many cases do you have? I have different samples approx.� 40-100

Can you match the pre and post measures within each case? I am not sure that I understood this question correctly, but if you mean: was the same student him/herself the same one for pre and post, then the answer is: No

How much attrition was there between the two measures? In terms of sample size? Not so big

What is the response scale on the items? In terms of St. deviation, it varies between: 0.67, 0.93 & 1.4 or 1.9. In terms of Max. and Min. 1 & 4 or 5.

It is possible that you can use a more elaborate kind of general linear model than the t-test, and use modified t-tested for post hoc comparisons, but a lot depends on the answers to the questions above. Please elaborate on this, I guess this is the ideal solution.

�

Just for clarifications, please allow me to debrief you with the following: In my research: I used an existing scale (many items included and this scale has in fact four sub-scales) and asked first year students (marketing major, business and accounting major). Also, I went to final year students in different majors (marketing, business and accounting) and gave them the same questions/items.

�

Finally, allow me to show how much I appreciate your kind help, I am really grateful. Thank you so much.

Samaa

�

--- On Sun, 7/26/09, Art Kendall [hidden email] wrote:

From: Art Kendall [hidden email]
Subject: Re: Scale or Items?
To: "Samaa Taher M. Attia" [hidden email], "SPSSX-L post" [hidden email]
Date: Sunday, July 26, 2009, 5:12 PM

Is this a single scale or are there 2 scales with different reliabilities? or is that pre and post?

Since the scales have poor reliability, do you have specific items you are interested in (i.e., have hypotheses for)?� Is there a subset of the items which does sum into a scale with better reliability?

Are there other variables you want to use in your model? Did you have a specific intervention, or is your interest strictly developmental?


How many cases do you have?

Can you match the pre and post measures within each case?
How much attrition was there between the two measures?

What is the response scale on the items?

It is possible that you can use a more elaborate kind of general linear model than the t-test, and use modified t-tested for post hoc comparisons, but a lot depends on the answers to the questions above.

Art Kendall
Social Research Consultants

Samaa Taher M. Attia wrote:

�

Dear Prof. Kendall,

Thank you so much for your kind help. Yes, the reliability is OK� 0.6 and 0.7, but what I am� asking is the data to address, can I use the items?

�

In� more details: I am comparing the communication apprehension between first and� final year, so I am using t test. But here, I want to use the communication apprehension items/data rather than the scale (as you know, we have oral and written). Can I use the items rather than the sacale? knowing that I did not find any study used the items before, all of them used scales.

�

Really, thank you so much and waiting for your kind reply,

Samaa

�

--- On Sun, 7/26/09, Art Kendall <Arthur.Kendall@...> wrote:
,

From: Art Kendall <Arthur.Kendall@...>
Subject: Re: Scale or Items?
To: "Stat-Help :)" <sataher@yahoo..com>
Date: Sunday, July 26, 2009, 3:04 PM

At a minimum you should use RELIABILITY to check that the syntax you are using to create the summative score implements the scoring key correctly.

What questions are you asking the data to address?


Art Kendall
Social Research Consultants

Stat-Help :) wrote:
> Hello all,
>
> This is my first contribution to this group, hope you can help me.
> I am doing a study on "Communication Apprehension", however, when I
> searched all previous studies, I found them used it as scales, but I
> would like to use it as items as well as scales. Is this possible, am
> I right?
>
> Your help is highly appreciated and thanks :))
>
> S.A

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD