Unfortunately this episode got a little out of control and is much too long. Therefore, I have decided to cut it in half and make an episode 4a and an episode 4b. I hope you like both episodes.

Episode 4a Inferential statistics – mind the gap

In episode 4a of talking texts inferential statistics – mind the gap We, Karina and Bernadette, discuss the text: “Inference and error in surveys” in: Survey Methodology by Groves et al. We explain the meaning of the words inference and error, summarize the text and talk about why it is relevant. In addition, we talk about Sherlock Holmes and how statistics can make the world a better place.

Time cues episode 4a

These time cues will help you find the topics we discussed or resume where you stopped listening.

00:00:00 Introduction of hosts, topic and text
00.02.34 Part 1
00.02.34 What does inference mean?
00:03:58 What is error in statistics?
00:04.36 How seeing error from a statistician’s point of view might improve our lives
00:05:43 Part 2 What is the text about and why does it matter?
00:05:59 Features of the text we found difficult
00:06:54 Summary of the text
00:08:50 Why is the text relevant

Episode 4b Inferential statistics – mind the gap

In episode 4b we explain some of the notions of error and statistical concepts. We then discuss one of the tasks from the text and present our solutions .

Time cues episode 4b

00:00:39 Part 3: different concepts in inferential statistics
00:01:09 Inferential and descriptive statistics
00:02:18 Error in studies
00:03:20 Errors of observation
00:04:58 Errors of non-observation
00:07:58 Part V: Discussion of Task from: Inference and Error in Surveys
00:08:26 Information before starting
00:09:19 What are the important parameters of the study?
00:10:37Discussion of task 1a)
00:11:24 Discussion of task 1b) and 1c)
00:15:49 Summary

Resources mentioned in the episode

Literature

GROVES, Robert M., et al. “Inference and error in surveys” in: Survey methodology. John Wiley & Sons, 2011. 39-67.

Further Links

Inference: https://www.youtube.com/watch?v=tFRXsngz4UQ

Errors: https://youtu.be/edzQQFNzFjM

Guests

Our guest in Episode 4a was Dr. Simona Helmsmüller. When we conducted the interview she was still Ms. Helmsmüller. You’ll find her here: https://www.h-brs.de/de/sv/simona-helmsmueller

Discussion of task 1

  1. These are the elements you have to keep in mind:
    Newspaper announcement that the sales of Personal digital assistants (PDA) have increased about 10% laptop and desktop pc sales have neither risen nor sunk. This claim is based on an on-line survey that was sent out using email addresses from five large internet providers:
    • Mode of collecting data self-administered email survey
    • Took place over a six-week period in 2002 from May 1st
    • 126,000 respondents, overall response rate was 13% (all in all 970.000)
    • 9.8% said they had bought a new PDA (i.e.12.000)
  1. a) Target population=U.S. adults 18 years and older Sample frame= subscribers to the 5 largest American Internet service providers (not everyone is a subscriber to an internet provider especially in .
    b) Coverage error: coverage error occurs when the sample frame does not represent all parts of the target population. Here the sample frame is subscibers to the 5 largest ISPs of the US. Therefore people who typically decide for services that are less mainstream will not be represented. Or people who do not use any email address at all. C) Suggestions: take samples not only from large internet providers but also from smaller internet providers and from email service providers you do not have to have internet at home to have an email address.
    b) Nonresponse error emerges when there is a discrepancy between the data provided by the respondents and the overall data. It occurs when the data for the sample is not fully provided. I.e. persons cannot be reached or refuse to answer. The  response rate for this study was 13% which means that roughly one out of 10 persons participated. The potential participants were contacted via email. Emails are often discarded or not opened if the sender is unknown. The amount of nonresponse error “If the nonresponse rate is very high, then the amount of the overestimation could be severe.” I really lack knowledge of a good response rate. Yet I assume as always it depends on the survey and sample. Still this response rate suggests that those who answered, had some kind of motivation. Therefore, this type of error is highly likely and that there might be some response bias involved. For why would someone do a survey on purchasing a certain good if they had not at least considered it? C) Suggestions: Send out several emails and remind people; Ask Internet service provider to send out the questionnaire.  
    b) Measurement error refers to a distorted result in the obtained data because the sample data deviates from the data obtained by measurement.
  2. As already implied above the motivation of those people who take part in the survey, is very likely to hinge on the fact that they at least considered purchasing the products. Therefore measurement error and response bias are very likely to occur. C) Include some kind of incentive like a prize or a voucher in order to motivate people to participate. C) suggestions: Advertise for the survey in order to increase people’s willingness to participate.

Quiz

  1. What kind of error is measurement error:
    1. Error of observation
    1. Error of non-observation
  2. Provide two kinds of error of observation
  3. Provide two kinds of error of non-observation
  4. What is the difference between target frame and sample frame?
  5. What is coverage-error?
  6. What is the difference between coverage error and sampling error?

Solutions: please download the pdf with the solutions here:

Credit:

Cover art: Anna Mahendra
Music: Felix Spieß