• Oct 5th 2012, 04:52 PM
faranakf
The following data gives the number of deaths due to accidental falls occurring in in the United States by Month
in 1970.
Month Number of deaths
Jan 1668
Feb 1407
Mar. 1370
Apr. 1309
May 1341
June 1338
July 1406
Aug. 1446
Sept. 1332
Oct. 1363
Nov. 1410
Dec 1526
(a) Using Stata or otherwise, calculate the expected number of deaths for each month if the probability of
accidental death is the same for each month. Report the results in a table.
(b) Is the data consistent with a uniform probability of death from accidental fall over time? Justify your answer
with the results from an appropriate statistical test.
(c) Is there evidence of a seasonal pattern of deaths? Explain why.
• Oct 5th 2012, 09:04 PM
chiro
Hey faranakf.

If you are calculating a proportion, then you need to supply what the data is in relation to (i.e. you have the deaths but how many people are considered each year? Is it the size of the total US population)?

Anyway getting the expectation of the number of deaths just uses the definition.

You should tell us what you've tried and what you're thinking.
• Oct 5th 2012, 10:05 PM
MaxJasper
These data follow a Poisson distribution based on following hypothesis test:

http://mathhelpforum.com/attachment....1&d=1349503427
• Oct 5th 2012, 11:09 PM
chiro
It is not wise to just find the "best fitted distribution" (also that p-value is rather low).

The distributions are based on assumptions and knowing the assumptions means knowing the process, and knowing the process (or trying to understand the real process) is what it's all about: if you can relate the process of your data in any way to the process of a known distribution derived from its assumptions in a convincing way, you are going to be way better off understanding and putting into context the realizations of your sample data.

Fitting stuff without any context at all is very dangerous.
• Oct 6th 2012, 12:08 AM
faranakf