Dear all,
I am trying to answer the question "how the retailers` acceptance of card payments affects the consumers card use".
First variable is "acceptance of card payments" (n=45; Yes/No), second is "card use among consumers" (n=89; Yes/No). Variables are taken from different data sets (surveys). As I understand, it is not feasible to calculate correlation in this case.
Would comparison of means be useful in this case ?
Could anyone please suggest which statistical test I could use to prove or disprove that card acceptance exert influence on card use among consumers (having only these two variables) ?
I apologise in advance if I used terms incorrectly.
As far as I know there is no way to deal with your problem. The only way is to collect new data with the 2 variables paired. Sorry, David Booth
As far as I know there is no way to deal with your problem. The only way is to collect new data with the 2 variables paired. Sorry, David Booth
Even worse than what David said: to be able to answer your questions you must randomize the consumers to the retailers. Thus, no purely observational data (e.g. from a survey) will be useful to answer your question, even if it comes from the same survey. Without randomization, the link between the retailers feature (e.g. acceptance of card payments) and the customers choice (e.g. card use) remains unresolveable, because there are too many possible (and plausible) links between a customers personality and her choice of a company. This is called confounding and can result in any amount of bias - and there is no chance to check or control this bias, exept by randomization. I think overlooking this fact (in observational studies / surveys) is a mayor reason why so many economic theories are not working. It's a recipe to produce bold claims from hot air. But, hey, that seems to be the way economics works...
Bogdan Borodynskyi
Yes. You can apply statistical technique to analyse the data. But you cannot go for differences in mean cause this is not a scale data. You have observations either YES or NO.
1) You can go for chi square = It will show any existing association between 'acceptance of card payments' & ' card use among consumers'.
2) You can go for binary logistic = it will show effect of 'card use among consumers'(Used as dichotomous independent variable) on 'acceptance of card payments' (Used as dichotomous dependent variable)
And you cannot go for correlation co-efficient. It is used for scale data like height, weight, inocme, age, soil Ph level etc.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS