Inference and Description
I thought that the work that you all did today was really rich, excellent stuff.
Part of what made it so rich was that it opened up so many issues and questions that are worth thinking more about. And, indirectly, Sarah, Jessica, Ken, and I were pushing you to the edges of some of those questions. A good example was when, tonight, Ken wanted to know what one group's hypothesis was and when he wouldn't settle for an answer that didn't have a firm hypothesis.
What's that all about? Why the focus on a hypothesis rather than just finding some stuff out?
One distinction that I have consistently found to be useful is between "description" and "inference." In statistics we talk about "descriptive statistics" and "inferential statistics." What is the difference? Is one a better thing to do than the other?
Think about the research that Jessica described with the oranges. She had some pretty low return rates for some of the orange releases. In some cases she had only one orange found in a particular location. What does a result like that show? What can you conclude from it? What are the limitations? What does this have to do with inference and description?
-- Bill
Part of what made it so rich was that it opened up so many issues and questions that are worth thinking more about. And, indirectly, Sarah, Jessica, Ken, and I were pushing you to the edges of some of those questions. A good example was when, tonight, Ken wanted to know what one group's hypothesis was and when he wouldn't settle for an answer that didn't have a firm hypothesis.
What's that all about? Why the focus on a hypothesis rather than just finding some stuff out?
One distinction that I have consistently found to be useful is between "description" and "inference." In statistics we talk about "descriptive statistics" and "inferential statistics." What is the difference? Is one a better thing to do than the other?
Think about the research that Jessica described with the oranges. She had some pretty low return rates for some of the orange releases. In some cases she had only one orange found in a particular location. What does a result like that show? What can you conclude from it? What are the limitations? What does this have to do with inference and description?
-- Bill
4 Comments:
In my mind, a hypothesis defines specific conditions under which your experiment is being performed. Without these specific parameters, it's much much harder to draw conclusions from the data you collect at the end. By demanding a hypothesis from a group, you're requiring specific parameters. Even if the group had specific parameters set, it's good to define what then you're looking for: it implies that you expect something specific too. "Finding some stuff out" well, you won't find anything out if you're not looking for something in particular. An experiment without enough controls isn't going to demonstrate a clear and specific relationship, so you won't be able to "find any stuff out."
Having said that, I would guess that descriptive statistics display data, but you can't necessarily infer anything from it. Inferential statistics would imply that a clear relationship can be seen when the data is looked at. Note: I don't actually know what either descriptive or inferential stats are.
When an experiment yields results that seem to have a lot of outliers, like Jessica's locations that only returned one orange, then it says, to me anyway, that the data point is in the realm of possibility, but not probability. That is, though it’s possible for something to happen, it’s less likely, maybe even unlikely enough to ignore. The limitation that needs to be accepted when ignoring those data points is that when you take a bigger sample, there might be more points that turn out to be similar to what seemed to be an outlier. If data does not show a clear relationship, but you draw inferential conclusions from it anyway, then there’s a good chance that a new data set will not support your conclusions. Data not showing a clear relationship is more descriptive as opposed to inferential.
By Natalie Jimenez, At January 12, 2007 3:34 PM
Natalie --
Your distinction between "probable" and "possible" is on the right track.
This is actually a pretty important distinction, and we are going to run into it head on this coming week.
SOMEBODY -- go to Wikipedia or some such place and look at up the terms inferential and descriptive statistics. Wikipedia goes into more detail than you need on the inferential side -- just try to get the main idea.
What does it mean to make an inference?
Natalie -- your group has taken samples at different elevations. Are you wanting to make some inferences?
-- Bill
By Bill Zoellick, At January 14, 2007 9:40 PM
Isnt an inference making a guess with the given data? Kinda like a whole educated guess kinda deal and you can make predictions based on it and the likely-hood of something actually happening. While descriptive statistics is something that summarizes and describes a set of data.
By Jordan Dolan, At January 15, 2007 2:07 PM
One of the reasons for having a firm hypothesis instead of 'finding some stuff out' is because with a firm hypothesis you have a more specified area of research. And without one, you could have a lot of information, but without specificity it is way harder to document, and shows little in depth information.
By Andrea Jarrett, At January 15, 2007 2:17 PM
Post a Comment
<< Home