[Master Thesis] Need help for a fitting research design

[Emerging]

Hi,

My advisor told me to look into and work out a fitting research design for my Master Thesis. I'm writing this about the effect of the corona pandemic on the sale of phishing tools on the dark web.
I'm going to analyze data from before the Corona-pandemic (to understand before the situation before the pandemic) and data after the initial beginning of the Corona-pandemic.

Can anyone give me any advice on a fitting research design? I was thinking about a retrospective study, however I don't think that is that common in these type of studies. Or the emperical cycle?

Thanks!

[timo]

Not sure I understand what you are asking for - I certainly don't know what you mean by empirical cycle or retrospective study. But to me, the general approach seems to be:

1) Define one or more quantitative measures for the sale of phishing tools, e.g. number of sales, volume of sales, number of different products offered, number of different products sold.

2) Find data source from which you can determine these measures. Note: The actual progress may be doing this step first and then defining measures that you have data for - it was just easier for me to describe the steps in this order.

3) Plot the measures over time in a suitable time binning (bonus points: with statistical error bars).

4) Define the time that you count as "Corona pandemic" and see if there are any visible trends in your graph.

4a) Alternatively, just try a few assumptions. E.g., for a suitable binning, fit two different constants to the data for the non-Corona and the Corona time intervals and check if these constants look significantly different (bonus points: calculate a statistical measure how different they are).

 

I imagine step 2 to be the hardest by far. Despite often planning to do so I have never tried to navigate around in the dark web. But the term already indicates that it will not be easy to get reliable overview data from it - especially since you are trying to monitor activities that are at least borderline illegal.

[ahmet]

1. I think I might help you via clarifying or specifying some instructions which given by @timo , or try to understand or follow these advices independehtly.

---->> you may use regression analysis. 

---->> try to find out whether there is a correlation between the variables you mention.

However, the first one might be easier than the second one. Because for the second one you will need numerical results.

 

You can also try to analyze curves. For instance you could try to find to which curve the new result (graph) fits (e.g: normal distribution, t- chi-square, etc.)

I think you can do this last one by approach,too (i.e. approximate curve) and this might ensure you analyze and comment more easily. 

 

I hope these helps. (As studiot states or implies  that he was dealing with applied mathematics ,I invite him to make external /additional advices.)

Edited October 26, 2020 by ahmet

[Emerging]

Hey! Thank you for the responses. Sorry if my question wasn't completely clear! It's for the social sciences to be more precise the master of business studies: finance and markets (perhaps this forum is a bit more exact oriented). I'll be handed a dataset, so that's a relief. 

 

For a research design I mean a design such as experimental research design, quasi-experimental etc. I will be looking at the past (pre-corona) and present (post initial start corona). and not e.g. looking at the future (prediction). Preditction requires a differnt type of research design. I hope that's a bit clearer.

 

@ahmetThank you! However I don't think regression is completely applicable here (though I might be wrong). I'll look into your suggestion about analyzing curves.

 Your suggestion at 4 and 4a is very helpful! In particular stating the timestamp of the "start" of the pandemic and look out for trends. @timo

 

[Ghideon]

2 minutes ago, Emerging said:

I'll be handed a dataset, so that's a relief. 

That's indeed a relief  

In addition to answers already given: My approach would be different depending on what data you have available. Probably I would start with:

-How is the data partitioned, for instance are statistics gathered globally or by region? This could have impact on the analysis since the impact of the pandemic did not hit everywhere simultaneously. 
-What kind of data is available, for instance if there are different categories of phishing tools available? This may allow for analysis of variations in available/likely targets during pandemic.
-Are there other data sets available for comparison? For instance reported number of phishing attempts before and during pandemic? 

You could try to find correlation (as stated by other members above) and/or , if data allows, look for cause-effect. Case study is also a possible design but less likely what you are looking for in this case.

The above is just a few notes intended as inspiration, feel free to ask follow up questions. 

[Emerging]

@Ghideon Thank you for the response as well. I think what I've been looking for is a kind of cause-effect research design. I agree that a case study is not such a good fit. I'm going to look into that!

Unfortunately I'm getting handed the data in a couple of weeks, which makes this even harder.. anyway thank you for the suggestions!

Edited October 26, 2020 by Emerging

[Prometheus]

A major consideration would be whether your data comes from a truly random sample - given the nature of what you describe, probably not. This will limit your options, but there are plenty of observational study designs you could utilise depending upon the exact nature of the study and what question you want to answer.