2. Did they "scrub" the data -- data hygiene?
Are you looking at all the data, or someone else's preconceptions of what
the data should look like?
3. Did they plot the data?
90% of what you conclude is usually visible in simple plots.
4. If there was a time series, was data plotted against time?
Always be on the look out for omitted variables -- time is number 1.
5. Were they looking for statistical significance against
an irrelevant value?
T-tests and p-values are meaningless unless you know what the hypotheses were.
Usually stat packages default to testing that the parameter is 0 and you may
not be interested in the default.
6. Were they R-squared challenged?
Has someone found a great R-squared then assumed that everything else was
OK because of that?
7. Is it clear how categorical variables have been coded?
Computer packages usually have a default categorical variable coding.
You need to know what it is to interpret the coefficients.
8. Did they do prediction out of sample? In-sample prediction may lead to a false sense of accuracy, and the problem gets worse as you increase the complexity of the model.