In my sample, the total number of selected firms is only 39 out of 700. I have selected them because they have publicly available annual reports containing the data I am looking for. However, many firms have their financial reports in an agency. I can collect their annual reports from this agency which can help me to add more firms to the sample, but these reports are unpublished. Can I combine the published and unpublished data then?
Please suggest me something.
Regards,
Md. Ali Rasel.
What kind of agency? And why are the unpublished reports there? Are they also public domain? If there is no bar on using them---then use them.
If you are saying that you have some sort of preliminary continuous-type data on all 700 firms, and final continuous-type data on just 39 of them - for various questions/dependent variables, one at a time - and you want to impute the preliminary data in place of the missing cases ... if that is your concern/question ... then you can probably do better. If x were for preliminary data, and y for final, for example, and a scatterplot were to appear linear through the origin, which often occurs with such establishment data, when dealing with continuous data, then you could 'predict'/impute/estimate for each 'missing' y-value. See the individual predictions (as opposed to totals, unless you want them also), in the following, for example: https://www.researchgate.net/publication/263036348_Properties_of_Weighted_Least_Squares_Regression_for_Cutoff_Sampling_in_Establishment_Surveys
The estimated square roots of the variances of the prediction errors - which are also impacted by bias - are often good indications of the accuracy for each predicted y-value. (In SAS PROC REG for example, this would be the STDI values outputted.)
But perhaps this is not on topic for what you meant to be asking.
Article Properties of Weighted Least Squares Regression for Cutoff S...
Dear Philip,
I am talking about a donor agency, which subsidises the firms for their business operation. However, the firms must submit their annual reports to this agency in order to gain further subsidy from it. These reports are not in public domain since the agency does not require them to be available publicly. I have to get its permission to access these reports, anyway. So, what do you think? should I use these unpublished reports?
Regards,
Md. Ali Rasel.
If you have explained to the agency, your purpose for accessing the reports, and still get permission for access, then I don't see a problem in using these reports; in the case of any bar to using them, I'd expect the agency to make you aware of this.
I also concur with Mr. Adams, since the source of the data is same with the published one, only that it was published. I suggest you can use both.
Published data are preferable because the data is already tested and peer-reviewed. Unpublished data may provide wrong result during simulation.
Ok, fair enough and thank you all for your prompt responses. More answers will be highly appreciated.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training