PLS and AMOS are used for testing the model and constructs. There are specific assumptions of each software. AMOS require data will be normally distributed, Large sample size >300, simple random sampling recommended, at least 3 items of each construct, however PLS has work better on small sample size i.e. 70, non-normal data, even 1 item of each construct and when model is much complex. 

On a smartPLS site you can find quite nice explanation under topic:

PLS-SEM Compared With CB-SEM

https://www.smartpls.com/documentation/pls-sem-compared-with-cb-sem

https://www.smartpls.com/documentation/learn-pls-sem-and-smartpls/pls-sem-compared-with-cbsem

Do not listen to the salesmen of a car. You need to listen to the automotive engineers themselves. So, in this case, you need to listen from the developer who developed the algorithm what they have to say regarding the analysis. They are statisticians, the expert in the field. I have read and summarized the information. Be an informed researcher yourself.

Thank you very much Prof. Zainudin Awang for your valuable and constructive comments. 

Also many thank to Mr. Konrad Kulikowski and Mr. Muhammad Anwar for your great comments 

Salam Ferdous bro, 

I am not promoting any software or approach here, hence I will answer your query with real facts that i have observed during using or reading about CB-sem, and VB-SEM. Your question was when to use these software. If you have reflective formative constructs or higher-order multidimensional constructs comprising reflective-formative constructs then you can only use warppls or smartpls. Amos can not handle such analysis. Another researcher as above said that pls can be used for small sample size, this is not a justified answer for using PLS. For PLS also, you need to have minimum 150 to 200 to have robust findings. The reason is, PLS results are consistent at large. thus, the more sample you have the more stable results will be. When you are only concerned with measurement model assessment with reflective formative constructs, the best and stable results u can only get from warppls. WarpPLS also calculates 6 model fit indices that have been published in top journals. Amos is better for simple models not complex one, amos is better with small numb of indicators. All comments that i gave you, are tested, verified........................ even I dare to ask very critical questions to developers just to know more abt using these approaches. All software u can use but before using try to understand your objective of the study, then if anything u find impossible in any software then go for another option. 

Walaikumussalam Brother Amir, 

Thank you very much for your constructive justifications. It will be helpful for me and others. One thing for WarpPLS  whats the sample range should be?

Use of software doesnot tell you the requirement for sample size. Try to calculate your sample size looking at your model. U can use Gpower also. Gpower or other sample size determination software will tell how much sample size would be minimum required to run a model. WarpPLS has some good advantages, which currently no software has, Thats true. But always go for more sample..... minimum 200 for SEM models. Low sample size is not the well justified answer for using PLS-SEM approach. 

Dear Dr Ferdous

I m totally agreed with Amir. For complex model/ higher order constructs u need either warp or Smart through justify your theory. However, every software has their own advantages and obviously u will be benefitted by those. Another thing u should know also proper uses of software. Thank you

Agreed Halima, Thanks for endorsing.

For my data analysis i need PLS license can anyone help me?

Naima Andleeb, WarpPLS has a 90 day trial period with full functionality of the software. Also, related to the above questions: WarpPLS will provide power analysis and estimate the minimum sample size needed bytaking into account the number of variables and factors that are in your model

In addition they say Warppls can handle non-linear SEM

in Warpls I can get scatterplot each relationship in the model to detect U curve, S curve and other. It can make descriptive statistic more deeply analysis

AMOS should be used when your constructs are REFLECTIVE in nature

PLS-SEM 3.0 works for both REFLECTIVE & FORMATIVE constructs with single item measurements and at higher levels as well. But in AMOS formative can't be done.

WARP-PLS works when you want to check moderation effect in categorical variables having more than 2 groups for e.g. Income level (level1, level2, level3, etc.) according to best of my knowledge.

However, some analysis still cannot be done in all the three like CCA