Showing posts with label charts. Show all posts
Showing posts with label charts. Show all posts

Sunday, August 9, 2020

Reporting Survey Data Using Maps

 In a previous post, I reported the value of reporting polling data using averages of many polls for the leading candidates. I mentioned that national polls are interesting but the US chooses a president based on the electoral college.

Because the electors are chosen by states, it makes sense to predict winners by observing how the population of a state is likely to vote and thereby decide how many electors "won" by a candidate. This assumes an elector does not go rogue and vote for their preferred candidate.

This map can of course change with every new poll, but has the potential for a more accurate prediction than charts of national polls.

The chart map video is from NBC News.

How do you create map charts?

Microsoft Excel has you covered-- see Create a Map chart in Excel.

  In Excel, you will find the map option labeled Geography under the Data tab.


You can download examples with several map charts from Microsoft.

Learn more about surveys in Creating Surveys available on AMAZON












Tuesday, July 28, 2020

Charting Dual Average Percentages as Linear Trends

An excellent example of presenting two sets of data from multiple sources over time can be found in the presentation of polling data on FiveThirtyEight.*

The point of  this post is to identify a useful way to present data from multiple sources over time.



The charts are continually updated as data from new polls are received*. The data for each of the two main candidates are plotted and a trend line shows the averages for each candidate. From left to right we see the progress based on the dates of each poll.

Notes about the chart and the data

1. Percentages can be averaged and yield a meaningful and easy way to interpret multiple sets of data.
2. The narrowing and widening of the trend lines offers a quick glance at what is happening for each candidate or data source.
3. Averaging polls from multiple sources helps avoid bias due to emphasizing preferred outcomes.
4. Below the chart are tables of data showing important information useful to research methods
   4.1 Dates are listed and are clearly important as events can change what happens.
   4.2 The data source is listed, which adds to confidence that we are not dealing with hidden data.
   4.3 The size of the sample is important to determine accuracy.
   4.4 The type of sample is important. So, LV = likely voters is more important to voting outcomes than is a general population sample. Other types of samples are listed.
   4.5 The results are listed along with the difference presented as a net result. On this day Biden was 8 points ahead of Trump.

It is worth noting in polls like this that events can change the trend. That is, just because Biden is consistently ahead of Trump does not mean this will be true later in the year. Also, it is important to remember that in the US, the popular vote does not determine who wins in a close election. As in 2016, the president won the electoral vote but lost the popular vote. Thus, polling based on electoral votes will be more useful for predicting election outcomes.

*Nate Silver is the founder of the website FiveThirtyEight. The story of Nate Silver's excellent predictions can be found in The Signal and the Noise: Why So Many Predictions Fail--but Some Don't

**Data Note that, in science writing, the word data takes a plural verb so, "data are" not "data is." To refer to a single unit of data, use the word datum. See dictionary.com In nonscience writing, data has been used as singular or plural.

If you are planning a survey project, please consider Creating Surveys on AMAZON

Cite this blog post

Sutton, G.W. (2020, July 28). Charting Dual Average Percentages as Linear Trends. Assessment, Statistics, & Research. https://statistics.suttong.com /2020/07/charting-dual-average-percentages-as.html



Links to Connections

My Page    www.suttong.com

  

My Books  AMAZON          and             GOOGLE STORE

 

FOLLOW   FACEBOOK   Geoff W. Sutton   TWITTER  @Geoff.W.Sutton

 

PINTEREST  www.pinterest.com/GeoffWSutton

 

Articles: Academia   Geoff W Sutton   ResearchGate   Geoffrey W Sutton 

 


Friday, July 17, 2020

How to Report Results of a Tracking Poll

Tracking polls are useful for evaluating changes in attitudes overtime. A simple yet effective approach is to plot the percentages of people responding in one of two ways each time you collect data.



In the example from the Washington Post-ABC News poll, the pollsters collected opinions of the public on the way the president was handling the coronavirus outbreak. By connecting the data points with different colored lines, the change is evident at a quick glance.

If you are interested in this particular subject, see the article by Clement and Balz, The Washington Post, July 17, 2020.


READ MORE about surveys and charts in Creating Surveys on AMAZON and other stores worldwide.







Links to Connections

My Page    www.suttong.com

  

My Books  AMAZON          and             GOOGLE STORE

 

FOLLOW   FACEBOOK   Geoff W. Sutton   TWITTER  @Geoff.W.Sutton

 

PINTEREST  www.pinterest.com/GeoffWSutton

 

Articles: Academia   Geoff W Sutton   ResearchGate   Geoffrey W Sutton 



Monday, June 15, 2020

Charting Police Shootings to death by Race and Year

The vertical bar chart is a useful method to show comparisons provided the data are accurate. The following chart presents data for three large major race groups in the US.




Chart from Statista for August 2020- 
See their page for current data, charts, and additional resources.


According to the US census, the estimated population of the US in 2019 was 328,239,523.
The estimated and rounded percentages of the major groups are:
  White 77%
  Black/ African American 13%
  Hispanic/Latino 18%

According to Statista.com, police shootings (to death) of Whites declined between 2017-2019, dropped then increased for Blacks, and for Hispanics.

Considering the small percentage of the population for Blacks and Hispanics, they fare worse than do Whites. See the statisca chart. Here's the link to the Statista chart.

Ad
Learn more about creating surveys, including calculating and reporting statistics in Creating Surveys on AMAZON    or   GOOGLE  Worldwide



Connections

My Page    www.suttong.com
  
My Books  AMAZON                       GOOGLE STORE

FACEBOOK   Geoff W. Sutton
TWITTER  @Geoff.W.Sutton

Publications (many free downloads)
 
Academia   Geoff W Sutton   (PhD)     

  ResearchGate   Geoffrey W Sutton   (PhD)



Tuesday, May 19, 2020

Weekly Progress US Death Rates Decline May15




We have an evident decrease on weekly deaths for the 7-days ending May 15 2020. Downward progress (green columns) resumed after the May 8 increase (yellow column).

Legend

M21 to A25 and M1 represent the Month and Day. The numbers below the dates are for the 7-days including the date. For example, M1 = 11,989 deaths in the 7-days before and including May 1, 2020. The first M dates are for March, then A for the April dates, then back to May again for M1 and so forth.

The data are beginning to look like a bell curve, but with many states allowing more freedom of movement, it is too early to tell if there will be a rise in a week or two.

The data are from


I use the download data file and create the chart in Excel.


Read more about statistics in these two books.


Creating Surveys on AMAZON










Read more about basic statistics in APPLIED STATISTICS: CONCEPTS FOR COUNSELORS at

AMAZON














Connections

Follow this blog

My Page    www.suttong.com

My Books  
 AMAZON     GOOGLE PLAY STORE

FACEBOOK  
 Geoff W. Sutton

TWITTER  @Geoff.W.Sutton

LinkedIN Geoffrey Sutton  PhD

Publications (many free downloads)
     
  Academia   Geoff W Sutton   (PhD)
     
  ResearchGate   Geoffrey W Sutton   (PhD)


Sunday, May 10, 2020

Covid Weekly Chart of US Deaths update





We have a slight increase on weekly deaths for the 7-days ending May 8 2020. Downward progress (green columns) has halted.


Legend
M21 to A25 and M1 represent the Monty and Day. The numbers below the dates are for the 7-days including the date. For example, M1 = 11,989 deaths in the 7-days before and including May 1, 2020. The first M dates are for March, then A for the April dates.

The data are beginning to look like a bell curve, but with many states allowing more freedom of movement, it is too early to tell if there will be a rise in a week or two.

The data are from


I use the download data file and create the chart in Excel.


Read more about statistics in these two books.


Creating Surveys on AMAZON












Read more about basic statistics in APPLIED STATISTICS: CONCEPTS FOR COUNSELORS at

AMAZON














Connections

Follow this blog

My Page    www.suttong.com

My Books  
 AMAZON     GOOGLE PLAY STORE

FACEBOOK  
 Geoff W. Sutton

TWITTER  @Geoff.W.Sutton

LinkedIN Geoffrey Sutton  PhD

Publications (many free downloads)
     
  Academia   Geoff W Sutton   (PhD)
     
  ResearchGate   Geoffrey W Sutton   (PhD)


Monday, April 27, 2020

Covid19 U S Deaths Weekly Chart



Are we there yet? The most recent weekly data suggest a downward move in the number of people who died in the past week.

In the previous chart, there appeared to be a channel or range between 1800 and 2000 per day with some anomalies. Now that we have more data, it is possible to group the data.

I am avoiding curves and means because it is not clear that there is a curve or that the data are normally distributed. The bar chart offers a clear picture of rapid increase and possibly (and hopefully a decline.

In this chart, I used weekly totals beginning with March 7, 2020 (M = March, A = April).

Important note: The numbers may be revised. This chart is for educational purposes only and not for planning.

Here is my source for the data  https://ourworldindata.org/grapher/total-daily-covid-deaths

Date               Deaths

M21 213
M28 1447
A4 5450
A11 11620
A18 18277
A25 13963







Read more about statistics in these two books.


Creating Surveys on AMAZON














Read more about basic statistics in APPLIED STATISTICS: CONCEPTS FOR COUNSELORS at

AMAZON














Connections

My Page    www.suttong.com

My Books  
 AMAZON     GOOGLE PLAY STORE

FACEBOOK  
 Geoff W. Sutton

TWITTER  @Geoff.W.Sutton

LinkedIN Geoffrey Sutton  PhD

Publications (many free downloads)
     
  Academia   Geoff W Sutton   (PhD)
     
  ResearchGate   Geoffrey W Sutton   (PhD)






















Tuesday, April 21, 2020

Charting Troubles and Covid-19




How will we know if opening up the country is safe?

My bottom line answer is when people stop dying from Covid-19.

But tracking actual deaths can help us know whether we are making progress or not.

We can measure progress against our own baseline. The chart provides an estimate of a baseline, which will need to be corrected when additional deaths are recorded.

Think of the baseline as a channel.

The national US baseline appears to be in the range of 1800 to 2000 with some outliers starting 8 April 2020. I am not using averages because the data are not obviously normalized. I prefer to look at a range of relatively stable values, which can be called a channel. The sharp deviations have to be ignored to get a sense of what is "typical" of a pattern. A new relatively stable range above or below this range should help us determine a new trend. If the numbers are corrected, we will need to revisit the range. I am looking at a move of about 20% either way for evidence of a new range.

Local baselines will be more informative.

I admit, that the suffering for many with Covid-19 looks terrible even when they survive. But there seems to be quite a variation in discomfort of survivors from near death pain and distress to "I didn't know I was infected."

Problem Charts- So Many Ways to Tell a Story

I am fed up with charts showing invalid comparisons, wildly hypothetical projections, and tables incorrectly comparing Covid-19 to other conditions.

But when I tried to grab the raw data and focus on how many of our fellow Americans actually died, I realized that even these data are estimates as you can see from the spikes representing people found on a certain day that had not been counted.

So, after the rapid ramp up, in March we see a rough baseline between 1800 and 2000 deaths per day between 8 April and yesterday, 20 April.

The arithmetic mean is not helpful given the outliers. The data are not normal so far, so forget standard deviations.

The curves referred to by scientists and repeated by news readers are not yet evident. There is a part of the data that curves upward until around April 6 or so.

Death is a lagging indicator as some have said. This makes obvious sense because we do not know if an infected person will survive or not.

Some charts extend lines into the future. That only makes sense if we have accurate past data and if the data we have establish a pattern.

Flatten the curve only makes sense if there is a curve. I would prefer "create a plunge" meaning that there is a plateau and we want to get down near zero in as straight a line as possible.

Some use log charts- I am not a fan when we do not know the trend.

Some charts plot moving averages. That's a good idea--as long as you understand the concept and use the best averaging method. For example, are you going to average 3-day, 5-days, or what?

Testing

Testing is important personally and locally but the national numbers are too iffy because we do not know how many people are tested with what type of test and the level of accuracy of such tests. (Tests are not perfect.)

Testing is most valuable to isolate infected people from loved ones, co-workers, and others. Testing will be even more helpful when there is treatment for the condition.

Testing can help identify "hot spots" needing immediate resources like a nursing home or workplace.

I support testing. But charts of confirmed cases are not useful until an entire population is tested.

And testing with different tests having different levels of accuracy interfere with making other than rough judgments.

Media Folk Science

I'd like to suggest quarantining so many social media posts but that's not going to happen. Best wishes on learning statistics and charts. It really can help you make wise decisions for your personal life as well as in schools and workplaces.

Check the data

Important note: The numbers may be revised. This chart is for educational purposes only and not for planning.

Here is my source for the data  https://ourworldindata.org/grapher/total-daily-covid-deaths

p.s. Data is a plural noun. Say and write "data are" NOT "data is."

1-Apr 909
2-Apr 1059
3-Apr 915
4-Apr 1104
5-Apr 1344
6-Apr 1146
7-Apr 1342
------
8-Apr
------
1906
9-Apr 1922
10-Apr 1873
11-Apr 2087
12-Apr 1831
13-Apr 1500
14-Apr 1541
15-Apr 2408
16-Apr 4928
17-Apr 2299
18-Apr 3770
19-Apr 1856
20-Apr 1772
21-Apr 1857

Added data not in the chart

              22-Apr                    2524
              23-Apr                    1721
              24-Apr                    3179    Median April 8-24 = 1906
              25-Apr                    1054
              26-Apr                    2172
              27-Apr                    1687


Read more about statistics in these two books.


Creating Surveys on AMAZON















Read more about basic statistics in APPLIED STATISTICS: CONCEPTS FOR COUNSELORS at

AMAZON














Connections

My Page    www.suttong.com
  
My Books  AMAZON                       GOOGLE STORE

FACEBOOK   Geoff W. Sutton
TWITTER  @Geoff.W.Sutton

Publications (many free downloads)
 
Academia   Geoff W Sutton   (PhD)     

  ResearchGate   Geoffrey W Sutton   (PhD)






Monday, October 30, 2017

Chart Example Marriage Age by year



The chart based on data from CDC 2015 provides an example of tracking three trends over time.

The bars indicate the percentage of births to unmarried women. The upper teal line represents the median age at first marriage and the orange broken line indicates median age at first birth.

Notice the "crossover" of the two lines referring to first birth and first marriage.

Note also the stabalized trend for births to unmarried women easily visible on the bar portion of the chart. About 40% of women are unmarried when their children are born.


You can read text related to the story at the BGSU weblink:
  https://www.bgsu.edu/ncfmr/resources/data/family-profiles/eickmeyer-payne-brown-manning-crossover-age-first-marriage-birth-fp-17-22.html


Creating Surveys
Available from AMAZON





Applied Statistics: Concepts for Counselors
Available from AMAZON













Connect
Geoffrey W. Sutton www.suttong.com
YouTube Psychology





Average Intelligence

  The concept of average intelligence is sometimes difficult to appreciate because the two words, average and intelligence, are sometimes no...