A note about the comments

A note about the comments

by Darlene Goldstein -
Number of replies: 0

hello everyone, In case it wasn't already clear........

In my comments I am trying to show where you can improve your report. Almost no one gets a note of at least 4 in the preliminary report. Don't worry about that !!

Your preliminary note is NOT indicative of your final report note - when modifying your report, you should focus on the substantive parts, those for which your input will make the most difference. And I will be more gentle in marking the final version :)

The part that seems to present the most difficulty is in writing the model mathematically. You should NOT use an R formula. You should state the model that you are fitting BEFORE you do the fitting. This will look something like:

y = b0 + b1 x1 + .... + e (<-- error)

If any explanatory / predictor variable is a FACTOR, then the number of coefficients will be the number of factor levels minus 1. For example, if there is a variable x that takes values low, medium, high, there should be 2 coefficients, one for 'medium' and one for 'high', and the interpretation of these is relative to the 'low' value. It doesn't matter whether these are coded 0, 1, 2 - you must treat the values as a FACTOR. In R, you can do this by:

variable.name.factor <- factor(variable.name)

for a variable called variable.name. See ?factor for the help.

You can check whether you did the coding correctly by looking at the associated df in the anova table - if it is 1 for a variable with more than 2 levels, you are treating the variable as numeric. For the low / medium / high example, the df should be 2 (ie, 3-1), not 1; etc......

So, in the mathematical / statistical part:

= state what model (or group of models) you are exploring / fitting (and why - why this class of models is appropriate for the problem)

= do the fitting

= CLEARLY explain how you are selecting a final model, and make sure that you define any terms. For example, if you are using AIC, the definition is not just Akaike Information Criterion, you need to say mathematically what this means, and how it is used to select a model (ie, smallest). You can use different criteria, but you need to EXPLAIN.

= do the model assessment - this generally involves diagnostics on (some type of) residuals

= write the final ESTIMATED model: this will look like y-hat = num0 + num1 x1 + ... (no error term), where num0, num1, etc. are numerical values for the coefficients

Several of you copy / pasted raw R output into your report. This is not permitted, and you cannot get an overall note higher than 5 (and maybe lower) if you do this. You can get formatted output by using (for example) the function xtable() in the R package xtable. There are also other ways to get R to format output, but I think for most purposes xtable should give you what you want.

You don't need to worry about page limitation, but don't go overboard - be clear and concise. Some of you did not follow the point size and margin restrictions - please don't do that. Using small point size or small margins makes it difficult for me to read your report. I will NOT be nice if you violate these in your final report.

And finally - please read again the report evaluation criteria and additional tips. These will help you to finalize your report.

Best regards,

Darlene