STATISTICS FROM A TO Z<br />-- CONFUSING CONCEPTS CLARIFIED
  • Home
    • Why This Book Is Needed
    • Articles List, Additional Concepts
    • Examples: 1-Page Summaries
    • Examples: Concept Flow Diagram
    • Examples: Compare and Contrast Tables
    • Examples: Cartoons
    • Example: Which to Use When Article
  • Buy
  • Blog
  • Sample Articles
  • Videos
  • Author
  • Communicate
  • Files
  • Errata

Statistics Tip of the Week: Always plot the data first. Calculated statistics alone can be misleading.

12/29/2016

1 Comment

 
It's been said that the first 3 laws of statistics are:
     1. Plot the data.
     2. Plot the data.
     3. Plot the data.
Others say the same thing using different words:
    1. Draw a picture.
    2. Draw a picture.
​    3. Draw a picture.
​
Let's say we have a couple of sets of (x,y) data, and we want to fit  Regression models to them. For Simple Linear Regression, we would first have to establish a Linear Correlation. If we just calculated the Correlation Coefficient, r, it would tell us that both the datasets graphed below have a very strong Linear Correlation.
Picture
We can see how that would apply for the data set pictured on the left. But  the one on the right is definitely not linear. We would be better off with non-linear regression, using a Polynomial model. 

This kind of thing happens more frequently than one might think. So, always plot the data first; calculated statistics alone can be misleading. And a picture of the data helps give you a more intuitive understanding of the data that you are analyzing. 
​
1 Comment

New Video: p, the p-value

12/18/2016

0 Comments

 
Picture
This new video: https://youtu.be/vyX4m89VkyI   
YouTube Channel for this book: http://bit.ly/2dD5H5f

This is the first of 5 videos planned for a playlist on 4 central concepts in Inferential Statistics:
  • p, the p-value
  • Alpha, the Level of Significance
  • Test Statistic
  • Critical Value
  • Alpha, p, Critical Value, and Test Statistic -- How They Work Together.
0 Comments

Statistics Tip of the Week: Power and the factors that affect it.

12/15/2016

6 Comments

 
Picture
In an Inferential Statistics test, a Beta Error is a "False Negative" (in contrast to an Alpha Error, which is a "False Positive"). Power is the Probability of not making a Beta Error. It is the Probability of correctly concluding that there is not a Statistically Significant difference, change, or effect. Put another way, Power is the Probability of accepting (Failing to Reject) the  Null Hypothesis, when the Null Hypothesis is true. 

Power is directly affected by Alpha, the Level of Significance, and n, the Sample Size. If we want to increase the Power of a test, the best way is to increase the Sample Size, n.

We could also increase Alpha. Alpha is the value we select as the maximum Probability of an Alpha Error which we will tolerate and still call the results "Statistically Significant". So, if we're willing to tolerate a higher Probability of an Alpha Error, we can reduce the Probability of a Beta Error. This is illustrated in my blog post on the Alpha Error and Beta Error see-saws.

As the two see-saws in the middle of the  graphic above demonstrate, Power has an inverse relationship with Effect Size (ES). If we want our test to be able to detect small Effect Sizes,  then we need to have a high value for Power. As we just said, we can get this by increasing the Sample Size, n, or increasing our tolerance for an Alpha Error.

The 2nd see-saw above shows that if we have low Power, then the detectable Effect Size will be high. 
6 Comments

You are not alone if you are confused by statistics #5

12/12/2016

0 Comments

 
Picture
I don't know what mathematicians think.  But I have been struck by inconsistencies in terminology and by disagreement about fundamental concepts among experts in statistics. 

For example, I was watching a video in which a professor kept mentioning "scatter". I had no idea what he was talking about, but eventually it became clear that "scatter" was the same thing as "variation" -- which is also known as "variability", "dispersion", and "spread". That's 5 terms for one concept.

Here's another example: in the equation y = f(x), y is variously known as the 
  • y variable
  • dependent variable
  • outcome variable
  • response variable
  • criterion variable
  • effect
Also, "SST" is the usual abbreviation for Sum of Squares Total. And 
SST = SSR (Sum of Squares Regression) + SSE (Sum of Squares Error)
But one author (at least) uses "SST" (Sum of Squares Treatment) instead of "SSR" -- which is very confusing. (I forget what they renamed Sum of Squares Total.)

And experts disagree about some very basic concepts. For example:
  • ​Some include stating the Alternative Hypothesis as a standard step in Hypothesis Testing. Others are "vehemently opposed" to its use.
  • Most insist that you must say "Fail to Reject" the Null Hypothesis. But some say that it's OK to "Accept" the Null Hypothesis.
  • And it seems generally accepted that overlapping Confidence Intervals means that there is no statistically significant difference. But some experts say that there can be an overlap of as much as 20% with a statistically significant difference.

0 Comments

Author talk at the Ridgefield Library

12/10/2016

0 Comments

 
Picture
Had my first author talk and book signing today -- at my hometown Ridgefield CT Library. Good turnout - 41 people. Lots of good questions. Signed and sold some books for Books on the Common. See the Files  page for the presentation slides.

0 Comments

Statistics Tip of the Week : the Binomial Shapeshifter

12/8/2016

0 Comments

 
The Binomial Distribution is used with Count data. It displays the Probabilities of Count data from Binomial Experiments. In a Binomial Experiment,
  • There are a fixed number of trials (e.g. coin flips)
  • Each trial can have only 1 of 2 outcomes.
  • The Probability of a given outcome is the same for each trial.
  • Each trial is Independent of the others

There are many Binomial Distributions. Each one is defined by a pair of  values for two Parameters, n and p. n is the number of trials, and p is the Probability of each trial.

The graphs below show the effect of varying n, while keeping the Probability the same at 50%. The Distribution retains its shape as n varies. But obviously, the Mean gets larger. 
Picture
The effect of varying the Probability, p, is more dramatic.
Picture
For small values of p, the bulk of the Distribution is heavier on the left. However, as described in my blog post of October 4, statistics describes this as being skewed to the right, that is, having a positive skew. (The skew is in the direction of the long tail.) For large values of p, the skew is to the left, because the bulk of the Distribution is on the right. 
0 Comments

Statistics Tip of the Week:  A comparison of various measures of Variation

12/1/2016

0 Comments

 
Variation is also known as "Variability", "Dispersion", "Spread", and "Scatter". (5 names for one thing is one more example why statistics is confusing.) Variation is 1 of 3 major categories of measures describing a Distribution or data set. The others are Center (aka "Central Tendency") with measures like Mean, Mode, and Median and Shape (with measures like Skew and Kurtosis). Variation measures how "spread out" the data is.
Picture
There are a number of different measures of Variation. This compare-and-contrast table shows the relative merits of each.
Picture

  • The Range is probably the least useful in statistics. It just tells you the highest and lowest values of a data set, and nothing about what's in between.
  • The Interquartile Range (IQR) can be quite useful for visualizing the distribution of the data and for comparing several data sets -- as described in my recent blog post.
  • Variance is the square of the Standard Deviation, and it is used as an interim step in the calculation of the latter. This squaring emphasizes the effects of very high or very low values. Another drawback is that it is in units of the data squared, which can be meaningless (e.g. square kilograms). There is a Chi-Square Test for the Variance, and Variances are used in F-tests and in the calculations in ANOVA.
  • The Mean Absolute Deviation is the average (unsquared) distance of the data points from the Mean. It is used when it is desirable to avoid overly emphasizing the effects of high and low values 
  • The Standard Deviation, being the square root of the Variance, does not overly emphasize the high and low values as the Variance does. Another major benefit is that it is in the same units as the data.
0 Comments

    Author

    Andrew A. (Andy) Jawlik is the author of the book, Statistics from A to Z -- Confusing Concepts Clarified, published by Wiley.

    Archives

    March 2021
    December 2020
    November 2020
    October 2020
    September 2020
    August 2020
    May 2020
    March 2020
    February 2020
    January 2020
    December 2019
    November 2019
    October 2019
    September 2019
    July 2019
    June 2019
    May 2019
    April 2019
    March 2019
    February 2019
    January 2019
    December 2018
    November 2018
    October 2018
    September 2018
    August 2018
    July 2018
    June 2018
    May 2018
    April 2018
    March 2018
    February 2018
    January 2018
    December 2017
    November 2017
    October 2017
    September 2017
    August 2017
    July 2017
    June 2017
    May 2017
    April 2017
    March 2017
    February 2017
    January 2017
    December 2016
    November 2016
    October 2016
    September 2016
    August 2016

    Categories

    All
    New Video
    Stats Tip Of The Week
    You Are Not Alone

    RSS Feed

  • Home
    • Why This Book Is Needed
    • Articles List, Additional Concepts
    • Examples: 1-Page Summaries
    • Examples: Concept Flow Diagram
    • Examples: Compare and Contrast Tables
    • Examples: Cartoons
    • Example: Which to Use When Article
  • Buy
  • Blog
  • Sample Articles
  • Videos
  • Author
  • Communicate
  • Files
  • Errata