Skip to main content

Watch those Covid 19 curves

Some highly intelligent scientists have used available data to plot trends related to the spread of Covid-19 such as number of identified infections, recoveries, and deaths. Others plotted ideas rather than data such as flattening the curve. Some rely on "common sense" and others rely on data.

Given my experience in teaching graduate research methods and statistics and talking with physicians and psychologists about statistics, I think it worth reminding any readers of this post that a lot of very intelligent people have difficulty with statistical models. There are many PhDs younger than I who know more methods of sophisticated data analyses than I learned during my PhD or since; however, not everyone stops to consider their assumptions when modeling data.

That's my point here--we need to review and challenge the assumptions about the data, the way the data are plotted, and how people interpret those data.

Missing data are very important to developing an accurate model. 
Unfortunately, in the early stages of a pandemic, there are a lot of missing data. Missing data can skew distributions. And it is very hard to identify outliers until sufficient data are available. People using different testing methods have different delay times in getting results. Some nations have more test kits in service than others.

Validity of data is crucial to developing a valid model. 
We have data from various nations, but we do not have evidence of verification of those data. Nor do we know if we are dealing with sufficient samples to make judgments about the population of an entire nation. Data from a small country can yield very different ratios than data from a large country. We need to know what percentage of a population are in the tested sample.

Generalization of data from one nation to another can be misleading.
In any research study, scientists usually caution readers about the limitations of their research. The current "real-time" reports of data translated into graphs showing different trends for different nations do not necessarily consider all the relevant variables that ensure trends from China may be applicable to future trends in the UK or the USA. So, in addition to the above problems, we cannot assume that the trend lines will be the same.

Different methods of plotting can lead to different conclusions.
I've seen plots of raw data and transformed data. Remember all that high school math about logarithms? Do you recall exponential growth curves? Plotting raw data is a good idea. Transforming those data requires that the assumptions for transformation make sense. It's one thing to present data in different ways as one seeks to understand the spread of a virus (or any other behavior), but it is quite another thing to make decisions for the public based on such assumptions.

Drawing conclusions from early data can lead to bad policy. 
It is generally wise to be conservative in drawing conclusions about data--especially when we do not know enough to truly plot future trends. I do not claim to know the best way to monitor the extent of the human tragedy due to Covid-19 virus effects. My inclination is to skeptically monitor the available data and various models, and follow conservative guidelines for the sake of my health and the health of others.

We might think about who has the most to gain from advice to do this or that in response to health and related economic policies during the time of the current pandemic and at other times as well. People usually promote ideas that are in their best interest (self-serving bias) and they view data and draw conclusions influenced by a confirmation bias (we ignore data that contradicts our assumptions).

In the case of Covid-19 policy, we might ask the following questions.

  1. Who will win or lose a future election based on the number of people who get infected or die?
  2. Who will gain or lose large sums of money based on the number of people who get infected or die?
  3. Who will win or lose a future election based on how well the economy is doing?
  4. Who will win or lose a future election based on the number of people who are unemployed?
  5. Who will gain or lose large sums of money based on the length of time people are under severe restrictions?
  6. Who will benefit from claiming shortages of various items considered "essential" or "nonessential"
  7. Who will benefit from having their employment labeled as "essential" or "nonessential?"
  8. How do decisions made during a crisis based on incomplete data reveal prejudice or even hate? For example, are the elderly worth less than the young? Are foreigners worth less than citizens? Are the lives of the wealthy worth more than the lives of the poor?
  9. What data tell us how policies will deferentially affect vulnerable populations? For example, I have hypothesized that requiring some people to remain at home might increase child abuse and domestic violence. Others have wondered about increased suicide in response to some healthcare policies focused heavily on Covid19.

You can probably generate more questions based on your profession and /or knowledge base.

About Math Literacy (Forbes)

My ads - I write and try to sell books. Download free samples from AMAZON.

Creating Surveys   AMAZON

Applied Statistics for Counselors     AMAZON


Popular posts from this blog

Personal Self-Concept Questionnaire (PSQ)

  The Personal Self-Concept Questionnaire  ( PSQ )   Overview The Personal Self-Concept Questionnaire (PSQ) measures self-concept based on ratings of 18 items, which are grouped into four categories: Self-fulfilment, autonomy, honesty, and emotional self-concept. Subscales : The PSQ has four subscales 1. Self-fulfilment (6 items) 2. Autonomy (4 items) 3. Honesty (3 items) 4. Emotional self-concept (5 items)  👉 [ Read more about Self-Concept and Self-Identity] The PSQ is a Likert-type scale with five response options ranging from totally disagree to totally agree. Reliability and Validity In the first study, coefficient alpha = .85 and in study two, alpha = .83. Data analysis supported a four-dimensional model (see the four categories above). Positive correlations with other self-concept measures were statistically significant. Other notes The authors estimated it took about 10 minutes to complete the PSQ. Their first study included people ages 12 to 36 ( n = 506). In the second s

Student Self-Efficacy

  Assessment name:  STUDENT SELF-EFFICACY SCALE * Note. This post has been updated to provide an available measure of student self-efficacy. ———- Scale overview:  The  student self-efficacy scale i s a 10-item measure of self-efficacy. It was developed using data from university nursing students in the United States. Authors: Melodie Rowbotham and Gerdamarie Schmitz Response Type:  A four-choice rating scale as follows: 1 = not at all true 2 = hardly true 3 = moderately true 4 = exactly true   Self-efficacy is the perception that a person can act in a way to achieve a desired goal.  Scale items There are 10 items. Examples: I am confident in my ability to learn, even if I am having a bad day. If I try hard enough, I can obtain the academic goals I desire.   Psychometric properties The authors reported that their sample scores ranged from 25 to 40 with a scale mean of 34.23 ( SD  = 3.80. Internal consistency was high at alpha = .84. The authors reported the results of a principal compon

Mathematics Self-Efficacy and Anxiety Questionnaire (MSEAQ)

  Scale name: Mathematics Self-Efficacy and Anxiety Questionnaire (MSEAQ) Scale overview: The Mathematics Self-Efficacy and Anxiety Questionnaire (MSEAQ) is a 29-item self-report measure of both mathematics self-efficacy and mathematics anxiety. Author: Diana Kathleen May Response Type: Items are rated on a 5-point Likert-type scale following a “no response” option: 1 = Never 2 = Seldom 3 = Sometimes 4 = Often 5 = usually Sample items 1. I feel confident enough to ask questions  in my mathematics class. 6. I worry that I will not be able to get a  good grade in my mathematics course.   Subscales and basic statistics for the MSEAQ       Self-Efficacy M = 44.11, SD = 10.78, alpha = .93       Anxiety M = 46.47, SD = 12.61, alpha = .93       Total Scale M = 90.58, SD = 22.78, alpha = .96 Reliability: See the Cronbach’s alpha levels reported above. Validity: There were significant positive correlations with similar measures. The results of a Fa