Urgent request. My suparvisor and my research model

[HussainWaasly]

Hi All,
My supervisor doesn't know any thing about PLS and he is usually asking me different questions about it.
At the end of my research he raised some questions. I answered him but he wants some references to prove that.

His concerns about new_6, new_7, and new_8 constructs.
In the survey I categorized the indicators to different sections for each above constructs.
For example new_6 has three indicators sections. Actually I can add three sub constructs to new_6 but the survey size will be big.

after analysis I removed some weak indicators based on PLS authors recommendations , so at the end I got one indicator for some section.
Now, he said you should have more than one indicator for each section.

I replied , i am testing the construct new_6 and I don't have sub constructs points to it , also the sections just a survey categorizing.
But he does not agree. Moreover, he said the construct new_8 has indicators more than new_6 and new_7 which is not ok.
In fact I don't know if his points are valid or no.

Please provide me your answers with references. I don't want to rewrite my research after 4 years of PhD study.

[Hengkov]

Hi,

First, I see your measurement model construct NEW 678 very poor, because many indicators not valid. I suggest, if you have large sample size, you check first outlier and drop out it for improve the outer model quality. And for negative loading, you must attend the question questionaire is right code or not.
About one or two indicators per constructs in PLS no problems, because identification model not effect (not same wtih AMOS, LISREL etc). But, may be if only one indicators, yuor cancepts construct cannot fully explain (cansistency at large), and I think your supervisor seeing it.

Best regards,

[ghozali]

Construct NEW _6, NEW_7 and NEW_8 have negative loading. You have to check the question whether there are negative wording. In this case the score is to be reversed. Droped the loading if it is below 0.70 (convergent validity)

[HussainWaasly]

Thanks for your useful inputs.
The above model is very old. i just put it here as an example for number of indicators not the result.

I have new model results with good and positive loading results and above 0.70.

My questions regarding number of indicators also if one construct has more number of indicators more than others.

Please provide me a references for your answers.
It will help me much . My supervisor hold my research and refused to submit it for VIVA until i answer him or change my model :(


Thanks again and BR

[Hengkov]

Hi,

Number of indicators related with conceptual and operational contructs/model. So, if you have 3-4 indicators per constructs 678 and the result outer model not good (many indicators not valid) and only one indicators is valid, this is mean your conceptual constructs very poor. Many factors effect this issues such outlier, respondent not understand questionaire, wrong code, etc. You must check all problem before deleted all items. If only used one indictor or single item, you cannot measure outer model (Validity and Reliability). This is mean, you cannot know the quality construct (not fully represented), so may be the result will no significance to test hypothesis. But, used single item in PLS no problem (identification model not effect, but always recommended to used many items-consistency at large). I hope this one article below will help you.

https://www.dropbox.com/s/b4jqlzzj1gs62 ... m.pdf?dl=0

Best regards,

[HussainWaasly]

Thanks again and again.
Sorry, for the below open items. i want to make sure i am in the right direction. so i can answer my Dr without any doubt .
Each construct has more than one indicators.
let me explain it more.
for example construct 6 , i have more than construct and all of them are OK.
the issue was in the servery design. i divided survey questionnaire for construct 6 to 3 categorizes (for example : A,B,C).
first category i named it for example 6A and first questionnaire i named it 6A1 (6: construct name, A: survey category, and 1 : number of questionnaire that belongs to the category).
so in the survey i'll have for construct 6 the following questionnaire ( 6A1,6A2,6A3,6B1,6B2,6B3,6C1,6C2,6C3).
after analysis and test-retest i found that the best indicators for example are ( 6A1,6B1,6C2,6C3)
So, the construct 6 will have 4 indicators.
Now, my Dr asked me why you have one questionnaire in the survey for section 6A and 6B ?
He wrote (The three constructs that measure impact have major issues. The items of the 6 construct has been poorly worded and many of it's variables have only one item. It is recommended to have at least four items measuring each variable. you need to justify in your thesis why he included only one item for some of the variables).

He wants to have 4 items for (6A,6B,6c)?
while i am measuring construct 6 not 6a,6B, and 6C that i don't represent them as a constructs in my model.

I hope the picture is very clear now?
Finally, my Dr analysis background is SPSS.

[Hengkov]

Hi,

I not fully understand your survey design. First, why your divided survey questionnaire for construct 6 into 3 category (A, B and C), your constructs is second-order or first order?
And three category it for same respondent or differ?
From your supervisor ask, I understand you have second-order construct. So for example:
Construct 6 have three dimensions say A, B, C.
Construct dimension A measure with three indicators and also for B and C too.
So, after analysis and test-retest you found the valid indicators for example are ( 6A1,6B1,6C2,6C3).
This is mean, your construct dimension "A" only measure with one item (6A1), for construct dimension "B" too (6B1), and for construct dimension 'C" just two items (6C2 and 6C3).
Your supervisor not agree, because each construct dimension have one items now (after analysis outer model). This is your result below:
A -> 6A1
B -> 6B1
C -> 6C2 and 6C3
Construct 6 -> ABC -> 6A1, 6B1, 6C2 and 6C3 (repeated indicators).
Using repeated indicators construct 6 have four indicators from three construct dimension.
If you draw in SmartPLS, each construct dimension only have one indicators ( in first order level) and second order level (construct 6) have four indicators.
So, you have problem about single indicators in construct dimension. And you must read my comment previous.

For addition:
You cannot put second-order construct directly in your model (with four indicators), without construct dimension. This is wrong, and reader not know construct 6 is second-order (but first order). This problem name misspesification model.

Best regards,

[HussainWaasly]

Hello,
It is reflective. i divided just for make it more clear and all for same respondents.
see the attached survey image.

Also, i am not measure A,B, C categories that is a different research topic and for type of respondent i have limitation to measure A,B, and C.
I am measuring the construct itself (the dimension in the attached survey image) .
Regards,

[Hengkov]

Hi,

Yes, you have second-order construct, so you must read my comments previous and also learn more about second-order constructs and hierarchical component model.

Regards,

[HussainWaasly]

Thanks again. Just small favor before reading second-order construct.
Could you plz drow just one construct based on your comments?
Regards