Basic Question on ‘Sampling Bias’

[Dhruba Lahiri]

 

Would welcome comments on the following problem

 

Suppose I ask a population of 2090 enterprises to choose between two different questions

Q1: A business should always strive for growth, profitability and innovation

OR

Q2: A business is about independence, acheiving personal satisfaction and enjoying your  work and lifestyle

Now out of a total response of 165 , 70 of them chose Q1 and another 95 chose Q2. For the purposes of subsequent analysis if I deliberately decided to rerstrict my analysis only to those who chose Q1 ( i.e the 70) and named them EO, would this constitute a case of sampling bias?

 

 

[Ingo Rohlfing]

This cannot be decided on the basis of the information that you provide. A sampling bias would be in place when the reason why some companies opt for Q1 is systematically related to your research question and hypothesis. So, what is your hypothesis?

[Dhruba Lahiri]

Thanks Ingo

The central hypothesis is that EO attributes (innovativeness, proactive and risk taking) and Absorptive Capacity (ACAP) together can explain a large part of the performance (measured in terms of potential value) of the firm. 

Look forward to your thoughts

 

[Ingo Rohlfing]

Well, that’s not my field of research and, as I said, the informal assessment of potential selection bias is one of substantive and theoretical reasoning. But I would say you do not have a problem because the independent variables should not be systematically related to one of the two types of businesses (e.g. risk taking should describe Q1 and Q2 businesses alike).

[Dhruba Lahiri]

Thanks Ingo

Your answer seems to substantiate my understanding on this matter.

Based on your earlier reply I delved into the matter a little bit more. As I understood based on readings there will be sampling bias if the dependent variable DV (in this case performance measured as potential Value) has different distribution between the two groups (Q1 Vs Q2). To investigate this I compared the means after splitting the file using an independent T-test and the results were non significant (p>0.05). Could this be an additional basis to argue that there is no sampling bias in choosing just the group who answered Q1 for my subsequent analysis?

Your insight would be highly helpful

[Ingo Rohlfing]

Yes, that’s useful auxiliary information in this context.

[Author]

Posts