2
$\begingroup$

I am new to the wonderland called Statistics. I studied a lot (finished Greene) about linear regression but I never got this clear in my head: if I am given data and asked to run a regression, what tests should I run to see whether data is appropriate to run a good regression? How do I know my data is good enough to give me a good model?

For example, we were asked to perform some data analysis on a survey our entire class conducted in Delhi. I didn't believe in the sampling strategies employed by other groups, so I was hesitant in performing a regression analysis. So, what should I do?

Improve this question
$\endgroup$
3
  • 5
    $\begingroup$ Regression can't indicate whether data were sampled well. That is a matter of considering the sample design and (as far as possible) comparing population values. Plotting data is the most obvious and by far the most useful preliminary to regression. $\endgroup$
    – Nick Cox
    Commented Dec 19, 2013 at 13:25
  • 1
    $\begingroup$ Why don't you believe the sampling strategies? If they are, in fact, bad, then GIGO applies (garbage in, garbage out) to regression analysis. But there may be ways to clean the garbage. $\endgroup$
    – Peter Flom
    Commented Dec 19, 2013 at 13:54
  • 1
    $\begingroup$ Is your concern just about the sampling? Or is it about assumptions of regression? Or how to set up a regression? Or what? $\endgroup$
    – Peter Flom
    Commented Dec 19, 2013 at 14:05

1 Answer 1

Reset to default
2
$\begingroup$

If you are given data and asked to run a regression, one question you should seek to answer is When Regression is not applicable? Without knowing your data and the type of the linear model you have been asked to fit it is only possible to get you started in this direction by providing some general advice. It may begin as follows:

There are many cases where you won't get useful results from regression. The two most common kinds of issues are (1) when your data contain major violations of regression assumptions and (2) when you don't have enough data (or of the right kinds). Core assumptions behind regression include

  • That there is in fact a relationship between the outcome variable and the predictor variables.
  • That observations are independent.
  • That the residuals are normally distributed and independent of the values of variables in the model.
  • That each predictor variable is not a linear combination of any others and is not extremely correlated with any others.
  • Additional assumptions depend on the nature of your dependent variable; for example ...

If this is what you are after, then please read more here. Of course, the time will come when you stop calling yourself a stats newbie, and will rightly disagree with some of the above, but those will be different questions or even answers.

Improve this answer
$\endgroup$
4
  • $\begingroup$ This doesn't seem to address the concerns about sampling. $\endgroup$
    – Peter Flom
    Commented Dec 19, 2013 at 13:55
  • 1
    $\begingroup$ This addresses the first para and the two questions therein. $\endgroup$
    – Hibernating
    Commented Dec 19, 2013 at 14:02
  • 1
    $\begingroup$ I think the question is a little unclear. "Data is good enough" is not clear. I didn't downvote your answer, I'm just wondering whether it is what lovekesh wants to know. $\endgroup$
    – Peter Flom
    Commented Dec 19, 2013 at 14:05
  • $\begingroup$ Well, thanks guys. I am really excited to get replied to my question. Let me rephrase myself. If i am a researcher and plan to do an econometric study, then i will clearly plan out my strategies about what variables would i choose and what sampling methods that would be. But in my case, our Batch (80 guys) divided into groups and sample 30 households per group. They were free to randomize as they wish in their respective area. I am comfortable with all the Regression diagnostics but what i want to know how would a econometrician would plan out the process till data collection ? $\endgroup$
    – lovekesh
    Commented Dec 20, 2013 at 7:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.