projecteverest

Proposal (Operational/other)

[Proposal: Operational/Other]: FarmEd All Countries - Analysing Qualitative Data - February 2018

by
Gabriel Raubenheimer
Gabriel Raubenheimer | Jan 25, 2018 | in FarmEd

I believe data is one of the most important tools we have, as an entrepreneurial organisation. By definition, we operate in an area of uncertainty. By definition, (good) data provides certainty. It is a phenomenally powerful tool in aiding our decision making, and it's excellent to see a FarmEd wide push toward collecting more data.

 

Our next step is to ask ourselves why we are collecting this data, and then let that inform how we collect it - beginning with the end in mind. Ultimately, we have to make our data useful. That is, the collection method should facilitate the usage. Here's how I propose we do that.

 

How are we using it?

For each piece of data we collect, we should have an application and/or potential application in mind. We should always ask why we are collecting this, and how else we might use it in the future.

 

For example, if we're doing a site analysis for a blueprint development, we might want to know:

  • Soil type and pH
  • Places on the site that are particularly shaded
  • The farmer's level of comfort with permaculture

Each of the pieces of information must be treated differently, because we'll use them differently.

  • Soil type and pH: both can be used by a computer to determine which crops can be planted, or to map common soil types in different areas of a country. 
  • Places on the site that are particularly shaded: used by the person creating the blueprint. Not used for data analysis, because there's not much chance of interpreting this in a statistically meaningful manner.
  • The farmer's level of comfort with permaculture: this may be important when designing a blueprint or engaging with the stakeholder. It will definitely be important in largescale data analysis and market analysis.

 

How do we allow it to be used in this way?

The next step is to facilitate the above, by considering how we can store this information. Continuing the previous example:

  • Soil type and pH: If we want to do any computer based analysis, we need the computer to recognised the soil type. That means always call sandy soil the same thing (ie 'SANDY SOIL'), so that the computer can just search for 'SANDY SOIL'. Loamy soil is ('LOAMY SOIL'), etc. We will make a standard operating procedure for this, so it's always the same. pH is easy, it's just a number.
  • Places on the site that are particularly shaded: We are unlikely to use this for data analysis - it's only relevant for that farm, and it only needs to be read and understand by the person designing the blueprint, not the computer. Therefore, it can just be text.
  • The farmer's level of comfort with permaculture: This is the most difficult and most interesting. On one hand, it's qualitative data, and often has to be to incorporate subtleties and nuances. On the other, it would be a very useful thing to do data analysis on. I propose the solution below.

 

 

The Curse of Qualitative (and the Solution)

Qualitative data is very important, as it is usually a better way to represent social dynamics than numbers are. So how can we retain this whilst allowing for deep analysis?

I suggest we develop a codeword system, where entries like the farmer's comfort above are always started with a specific summative keyword. Using that example, the farmer may feel excited about the use of permaculture, but may also be slightly concerned about buying more seeds. On balance, this is judged as positive, so the entry begins with 'PSTV'. This way, when doing analysis, the computer can just search for the words 'PSTV', 'NGTV', 'NTRL', etc. It will be important to make sure these words don't come up elsewhere in that field, but a good SOP will facilitate that.

 

Action points

We are close to finishing an initial framework for the database, and will be developing this SOP tomorrow. It will need iteration, but it is important that we start collecting data like this as soon as possible.

If you have any ideas, or have a way to improve it, it will be important to continue iterate and refine it, so please drop them below!

edited on 6th September 2018, 01:09 by Justin Hakeem

Corey Middleton Jan 31, 2018

This is essential. the qualitative data naming system is crucial, you are correct. I know for a fact that many agencies have naming metrics for qualitative data.

The following is a suggestion you may have already thought of, but let me know your thoughts.

the way that I have seen it done is via excel spreadsheets which are coded to produce an automatically generated name based on a series of drop-down selections and comma separated values.

For instance, you may have:

Qualitative

series 1 (mood) Pos, Neg, Neut
Series 2 (Topic) Perma, CompCrop, Monocrop

Quantitative
Series 1 (crop type) Crt, Csva, Lttc

And so on.
at the end, you have coded the sheet to produce a unique name for the information relevant to that individual. the name is a collection of comma separated values Which means you are able to index the database quite easily.

It wouldn't take a whole lot of time to code the excel sheet, but the number of data points could end up being enormous. I can show an example if need be?

Users tagged:

Reply 1

Corey Middleton Jan 31, 2018

also ciaran

Users tagged:

Reply 0

Gabriel Raubenheimer Feb 1, 2018

Absolutely, this goes back to reducing human error, and making the data input experience as efficient and easy as possible.

If I understand correctly, you mean being able to select a crop, for example, from a drop-down, and have the code and other information for it appear automatically? If you have an example, it would be excellent to see.

That would make an excellent front-end for the back-end database, because it will display information very nicely.

The other way we can action it is when inputting via Forms, by, as you say, making a series of drop-downs rather than free-form entries. If it needs to be put into Sheets, we can do that automatically too, via Zapier.

Do you have any more information/links regarding how agencies are making codes for qualitative information?

Reply 0

Corey Middleton Feb 1, 2018

What is your email Gab? I can forward you an example?

Reply 0

Corey Middleton Feb 1, 2018

All good, got it

Reply 0

View all replies (4)

Wade Tink 9 months ago

Status label added: Proposal (Operational/other)

Results of this proposal? Was it adopted? Current status

Reply 0

Wade Tink 9 months ago

Was this adopted? Do we have a current status? Suggest a new post highlighting results/operational changes or we can move this to adopted status.

Users tagged:

Reply 0

Share