I am using Panel data and my model is mostly dependent on Secondary data, but my university insisted that I must also have a primary tool. At what time do I merge the outcome of the secondary data and the outcome of the primary data?
Hi Robert,
It is not clear to me how and why you would need to collect and merge primary data with your retrospective existing secondary data?
Let's say that you have 3 waves of panel data that were collected over the course of the past 3 years. Then how exactly are you expected to collect primary data and then merge them with the existing data? Would you sample from the subjects for whom you have 3 years of data? Would you interview/collect new data on all subjects? If so, how would you treat those subjects with incomplete data? Would you ignore them? Lastly, what sort of primary data would you be collecting on these subjects, that would complement the existing data? The only thing I can think of is that you are expected to conduct some sort of qualitative analysis, but you did not explicitly say as much.
There is something missing from your post that limits the ability to provide a meaningful response. Please clarify exactly what data you have already collected, the sample size of those subjects over how how many waves, and what additional data are expected to collect, and how?
Ariel
GENERAL PRACTICE: Generally, we do not combine primary and secondary data. However, there is always an exception, if the model requires an adjustment by using secondary data, i.e. a firm's condition that is affected by market condition. In such a case, your firm condition is a primary data. This firm condition, i.e. financial performance (primary data), is affected by the market condition (hence secondary data).
EXAMPLE: In one study (see attached article) on cash flow modeling in construction industry. the cash flow of the firm is a primary data. The model is constructed with the assumption that the firm's cash flow depends on the firms' own expected cash flow (primary data) and the effect of the market condition (secondary data). See equation (1) in attached article. In such a case, we do not merge, we combine them as individual components in the structural equation.
Another example in integrating primary and secondary data in a model is SWOT analysis. Strength and weakness is internal, i.e. primary data. Threat and opportunities are external to the firm (but related to the firm)---this is secondary data, i.e. industry, economy, market, etc.
THESIS ADVISORY: It is a common practice for us to advise thesis students to use primary data (collected through survey or using company data) and integrate with data from secondary source, i.e. IMF, World Bank, SP500, etc. We do not use the term "merging", but we say making full use of primary and secondary data---the purpose is to make the student's vision more expansive and the construction of the model more integrated (using primary and secondary data).
REFERENCE: See attached articles and citation therein for more cases. Cheers.
It is hard to combine the two unless ( in special circumstances) you Identify a gap that primary data can complement to complete it!...so do a gap-analysis!...but results will be analyzed separately then intuition is inferred.
Its not a common practice to merge primary data with secondary one. However, one can run a secondary data get the result and seek for primary data(public opinion) to buttress or otherwise the result from the secondary data but can not be merge. In some non-parametric analysis where the work entails for instance running a qualitative data (binary) on some quantitative ones such as in Logit/probit models.
You need to analyse both separately, statistically test them separately, analyse the results separately. But compare the two analyse and comment logically on each separately, compare and contrast both in result discussions. For example the primary data will make your work richer because it is feeling the pulse of the respondents stakeholders directly. Primary data is current and captures the research environment better.
In fact it is not recommended to use primary and secondary data to justify the hypotheses. But, a researcher may use Quantitative and Qualitative research if needed to justify the findings (Mixed Method).
I read from top to down. What a great answer from all experts. I am thinking about the same issue. I have developed a demand model for product Z by utilizing the theory of demand. Based on the theory, some indicators are from secondary data such as quantity demand, income, price, price of substitution products and all these data are the time-series data.
On the other hand, some factors such as consumer behaviour cannot directly measure. I called it a construct. I am using the theory of planned behaviour to develop the construct. Then, collect the data thru a questionnaire via primary data. Can anybody suggest to me, how to link these two types of analysis? One from the time-series data side; and another from the primary data side? Tq.
- Badges
- Science topic