EDA

Exploratory Data Analysis also known as EDA is a way to dive into your data and try to understand it, learn from it and use it to help drive forward. Some of the basic things you can get from this are: what is the mean or average of the data, how many samples are there, how many missing data are there, what is the median, mode, standard deviation, variance, coefficient of variation, interquartile range, Q1, Q3, min, max, etc.

Why would you want to do this?

It's a way to gain knowledge of all the data that you have been collecting and it can be done altogether, broken up into various domains, lithologies, by variable, etc. Understanding patterns, relationships and characteristics of datasets is needed before diving in and doing further complex analysis and model building. Key assumptions have to be checked, further work might be needed, and it can help drive what methodologies and decisions are made down the road throughout the estimation and other geostatistical workflows.

Whenever I'm trying to teach someone about something that I know about (or I think I know LOL) I try to relate it to something they know. I've been told that I'm great at simplifying complex problems into details that can be understood by most audiences. Don't forget, someday I want to be a professor...that is in my long-term plan once I have all of these fascinating stories to relate mathematics, statistics, geostatistics back to. I love to teach but more importantly, I love it when someone that tells me they won't ever be able to understand it, finally understands. That is a game for me, that's what excites me, makes me happy is when someone is winning...I'm winning. I try various different ways with words, visuals, stories, etc. to break it down and try to make it relatable for them as well as in the form they like to learn in. Anyways getting off track.

So, what can I relate this to. Let's relate it to buying a car. Most people in their lives will buy a car at some point, even if you are like me and try to constantly make your brother, dad or husband do it for you...they always want you to learn how to do it, so you have to do it. And oh, when I'm driving through the parking lot and see the salesman start walking toward my car as I'm increasing speed trying to get away from them because I don't want to talk to them LOL! They always try to wheel and deal. And I know that you all have done the same thing! This week, I went and looked at a Chevy Traverse and this guy named Chewy came up to me trying to shmooze me stating "Hey do you work here"...Really? That isn't going to work dude. "Oh yeah we have lots of Traverse" and they only had three. Then he said "oh I've only been here a week" now I used to say the same thing when I was giving tours at Montana Tech because I knew 1) it made them laugh when I told them actually, I had been doing it for a year or 2) I got better scores since they felt I did a great job being so new. Anyways, back to this.

When you are buying a car, what sort of analysis are you doing? You first are browsing all the vehicles; you should sort of have in mind what you want but maybe you don't. You have colors you like and ones you really don't like. You may need a bigger car for your family or maybe you just want a nice simple commuter that has great gas mileage. You look at the car - is it new or used? How many miles does it have? What sort of miles to the gallon does it get on highway and in town? What safety features does it have? Was it owned by someone previous? Was it a rental previously? Is there anything missing? (sometimes those car lots like to hide things---my 2nd car a Pontiac Grand Am had been wrecked prior and the bracket that holds the lights was broken). Are there any dents? scratches? or mechanical issues? So just like data you are cautious with data that has significant quality issues.

When looking at fuel efficiency, engine size, mileage and safety ratings, this is similar to exploring data for distributions and spread of values or to see what the typical values are. Does the car suit your needs? Is the data in a format that suits your needs?

When looking at a car you may look at relationships like the trade-off between fuel efficiency and engine power. In EDA you look at relationships among variables and how or if they impact each other - any dependencies in data.

When buying a car it may lead to further hypotheses like "A car with higher mileage should be cheaper" or "Toyotas can have infinite mileage and last forever" (no, just me? ha ha my car battery lasted 10 years...please don't try that at home!). EDA also does similar allowing you to generate insights and further theories about your data building further steps, analysis and model building.

Before finalizing the purchase on a car, you need to make sure the car fits your requirements much like you need to ensure that the data is ready for modeling. You may need to "tune" the data by transforming, further domaining, sending it back, or cleaning it, similar to prepping a car for optimal use.

In summary whether you are buying a car or building a model, ensure that you use EDA to ensure quality, understanding all the features and prepare for making well-informed decisions. Don't end up like I did with my broken headlight bracket that eventually caused my blinker to emit the noise erratically on a trip from Butte, MT to Seattle, WA. At first, I thought I accidentally left on my blinker but when the sound would speed up or slow down and be intermittent, that was when I knew there was something wrong.

Next week, we'll talk about different types and techniques for EDA, what it means and what to look for. I find that throughout my career, sometimes I would walk around and see different employees run EDA but just as a checkbox exercise and maybe that was because they had a historic deposit that was set in stone, or they had a recipe, or they were told to do it or they didn't really know what they needed it for. So, next week, let's take a look and see how we can use this information for our benefit and to push us further down the road.

#WisdomWednesdayWithCW #Geostatistics