Hi Daniel,

To me statistical analysis is a "black box" that contain series of processes & mechanisms to dissect all the data collected in order to provide useful information to researchers or help them to make informed decisions. Statistical analysis is analogous like a “Statistic Factory” that process raw data into well informed output that a researcher can make use of it for decision making / conclusion. Nowadays, we might not need to know the formulas like last time but we need to understand what the statistical analysis serves and how to interpret the output it has generated.

Statistical analysis is part of data analysis. Hence, understanding the data attributes are important as they are the starting point to understand statistical analysis e.g. data related to concepts / constructs in a framework / model, data types (nominal, ordinal, interval, ratio), sampling, population etc. Because data attributes determine what statistical tests / modeling to be used e.g. parametric, non-parametric tests, covariance-based SEM, variance-based SEM etc.

Within each statistical analysis, further analogy can be used e.g. exploratory factor analysis is analogous like a tool to group / summarize all the raw data into higher order groups / factors. I think you can also refer to one of Andy Field famous books called “Discovering Statistics Using SPSS” that contains some analogies that he used to clarify certain terms in statistics. All the best.

Regards,

Fung

Statistical analysis is a tool to achieve something, usually to find out about a situation (descriptive) or a relationship (Inferential). You can visualise this with diagrams, models or even images, but it is important that the labelling used is accurate, and reflects the data analysis that you are using. I have seen photographs used very effectively to illustrate for example regression analysis, which establishes the influence the input variables have on the output variable. E.g.  sunshine + rain + good soil influence plant growth, all of which can be depicted in pictures.

I find that graphic displays are the most effective way of presenting complicated statistical results to both lay audiences and those with scientific backgrounds (in fact, I usually graph results out to better my own understanding of the data and results). There are several excellent books on graphing data, and every good statistical software package has a vast array of graphing tools. I suggest that you spend some time considering which display best presents your data and think of how a lay audience would view these results.

Thanks for your help everyone! Indeed, charting is important in clarifying the concepts. I had a meeting with some senior management in the industry and I was given 15 minutes to present my analysis over a phone-call. I was faced with sheer amount of pressure to bring across what I had found. Of course, I gave what they wanted and that was the insight impacting P/L but avoiding the jargons and explaining the technique over a phone-call when I was asked to explain was really the challenge. I had a followup email with them, explaining the technique in very simple terms - to the point where a child can understand (hopefully..) - and that was helpful though.

I had my fair share in the academic and I taught my students by using graphs. But I do use everyday items or some games that the youngsters play to relate the techniques to them - to make it interesting! But I have been struggling to find all possible simplifications to help me explain, without losing the quality and substance and at the same time fool-proof.

Sounds like you're on the right track, Daniel. Good luck!

Thanks Ariel!

For many applications of statistical analysis, I think the core concept is "explaining variation." In other words, we want to understand the sources of different outcomes.

Even simple bar charts are often constructed for that purposes, so I think your emphasis on communicating through those tools could indeed be useful.

Daniel, you asked a very important question. Recently I have come across similar problem - I have studied for some time to be able to apply some statistical techniques, and now I realized that the things, which are clear to me, are totally unclear to my colleagues. And it is really not easy to explain it to them. Sometimes it feels really strange. I never wanted to be a teacher but now I have to find the teacher in me. I am trying some simple examples and then build them up into more complex examples. I hope it works. Good luck to you!

I guess the real difficulty lies in transforming the mathematical expression into our daily activities. It is quite straightforward to explain anomalies in very lay-man's term: there is a large group of people in the room and a handful who is influential are outside that group. This handful of people are able to cause the main group of people in the room to move towards them progressively. This handful group of people are anomalies and the larger group of people refers to us. I'm still refining this example though, so please bear with me. I explained this concept to a group of year 11/12 students and they understand it. I also used billiard balls to explain cluster analysis!

Graphic displays are the way to go. Even simple scatter plots are helpful to display clustered behaviours, although much more complex relationships can be very nicely displayed using the many graphing tool available in most statistical packages (and stand alone packages as well).

I think one of the best vehicles for demonstrating clusters is (paradoxically) the qualitative analysis software NVivo. With one simple function (production of a tag/word cloud) the alternative views available to see are tag/word cloud, cluster analysis and 3D multiple correspondence analysis - the latter can also be animated and rotated. In a simple 10 minute demonstration of this to graduate, postgraduate and doctoral students they can quickly and easily grasp concepts of clustering (and even factor analysis) - I always employ this approach on postgrad and doctoral research methods teaching

I believe that every major stats software program has graphing capabilities for cluster data. Generally it falls under dendrograms or some other nomenclature.