Analyzing Pooled Likert Data

Hi Everyone,

I'm analyzing Likert Survey Responses for a question "How satisfied are you with your Item?" (1 = Very Dissatisfied / 5 = Very Satisfied). Here's example survey data:

1 2 3 4 5 n % Very Satisfied
Item A 2 5 6 15 22 50 44%
Item B 1 0 1 0 8 10 80%
Item C 1 1 0 3 7 12 58%
Item D 0 1 2 2 5 10 50%

Item A Very Satisfied = 22 (44%) / Other Items (B-D) Very Satisfied = 20 (63%)

Here's my dilema - Items B-D have a low individual sample sizes (< 11) but when pooled have a sample size > 30 (32). Due to the low samples sizes for Items B-D I would not be comfortable stating that fewer respondents are very satisfied with Item A than Items B-D individually. When Items B-D are pooled the total sample size increases > 30 however; I'm having a tough time acknowledging that fewer respondents are Very Satisfied with Item A than Items B-D collectively. In other words Item A lags behind the rest.

So my question is can we pool survey responses and make inferences between the pooled samples and another sample when the individual pooled sample sizes are low? What would be the reasons for/against doing this?

Many Thanks in Advance....



TS Contributor
I do not know if I will be able to answer your question; but I would like to preliminarily know (i) how many items do you have (4 in total?); (ii) what inference you wish to make (i.e., what's your goal)?


I would be pooling 3 items (B,C,D) and then comparing them to A (4 items Total)
The inference would then be; is A <> (B,C,D)