The Student Room Group

Finding the right statistical test to run

Hi
I am running a test on a set of data about birds and the time of day they are feeding at a feeding station - I have the data but I am unsure the best way to arrange the data so I can find out if there is any significant variation between the 3 times 07.00-10.59, 11.00-14.59 and 15.00-18.59 and number of visits. Also what stats test to run in minitab. I have attached my excel spreadsheet.

The original data is rows A-E on Sheet 1, the rest is my attempt at working out how to sort out this data into a format I can run through minitab. I tired attaching my mini tab sheet but it keeps saying invalid type.

I would greatly appreciate any advice and if any additional information is required I will gladly supply it.

Kevin
Original post by Greenpegasus
Hi
I am running a test on a set of data about birds and the time of day they are feeding at a feeding station - I have the data but I am unsure the best way to arrange the data so I can find out if there is any significant variation between the 3 times 07.00-10.59, 11.00-14.59 and 15.00-18.59 and number of visits. Also what stats test to run in minitab. I have attached my excel spreadsheet.

The original data is rows A-E on Sheet 1, the rest is my attempt at working out how to sort out this data into a format I can run through minitab. I tired attaching my mini tab sheet but it keeps saying invalid type.

I would greatly appreciate any advice and if any additional information is required I will gladly supply it.

Kevin


The simplest analysis to perform would probably be to do a linear regression with the number of number of visits as the outcome variable and with the period of the day as the only explanatory variable (i.e. a three level categorical variable A=first time period, B= second, C=third).

If you do this then you find that mean number of feeds for time period "C" is less than for "A" (p = 0.04) and that for "B" is less than for "A" (p = 0.62). Unfortunately, the overall p-value for the regression is 0.104 (indicating that the period of the day variable is not associated with the number of feeds).

This pattern is repeated if you do a series of t-tests: C versus not-C (p= 0.01); C versus A (p = 0.04); C versus B (0.08); B versus A (p = 0.66).

What this is telling me is that had you set up your initial hypothesis as:

H0: number of feeding visits for period C is the same as for the the others taken together.

then you could have rejected H0. However, if you set up

H0: number of feeding visits is independent of time period.

then you could not reject H0.

I would also note that there appears to be a secular effect in the data - the number of feeding visits seems to climb over time. Is this a problem that needs to be taken account of in the analysis?
Reply 2
Thank you for the thorough evaluation of the data I provided, the problem with this data is it was independently gathered and I have to make sense of it, week 3 sees a significant spike of visits, I am not sure if this is down to an increase in food used, a cold snap in the weather forcing the birds to come into urban areas, the disappearance of a predator, I don't have adequate information, thankfully this is a hypothetical situation so I can perhaps find adequate answers.

weeks 1, 2 and 3 are very similar and week 3 being the significant difference. So yes the number of visits climbs into week 3, then levels to that similar to weeks 1 and 2 by week 4, this will be taken into account of the analysis.
Reply 3
I have also been informed all my statistical tests have to be run through minitab, not sure why this is the case.
Original post by Greenpegasus
Thank you for the thorough evaluation of the data I provided, the problem with this data is it was independently gathered and I have to make sense of it,


In the circumstances I think the best that you can argue for is that there is weak evidence that time period C sees less feeding activity. The evidence is of sufficient quality to justify (i.e. could go on a grant application form!) a further more precise study, but probably no more than that.

It's one of the key rules in setting up an experiment that the hypotheses must be pre-specified - otherwise you can easily come up with "significant" results by data dredging - and these results will not reproduce in repeat experimentation.
Original post by Greenpegasus
I have also been informed all my statistical tests have to be run through minitab, not sure why this is the case.


Odd. Minitab's largely gone out of fashion these days. I don't know any statisticians who use it. still, it should provide the faciliies you need.
Reply 6
Original post by Gregorius
Odd. Minitab's largely gone out of fashion these days. I don't know any statisticians who use it. still, it should provide the faciliies you need.


Yeah I am not 100% comfortable with it, my lecturer had said he wanted all pupils to be comfortable using both excel and mini tab.

My hypothesis was, "The time of dayis directly associated with the feeding pattern of birds at feeding stations" not sure the strength of the hypothesis or whether I should reject it for an alternative such as "The time of day has no significance when birds feed at feeding stations"
Original post by Greenpegasus
Yeah I am not 100% comfortable with it, my lecturer had said he wanted all pupils to be comfortable using both excel and mini tab.


I sincerely hope your lecturer was not suggesting Excel for statistical analysis! It's something of a joke among statisticians. FIne for data prep though.

My hypothesis was, "The time of dayis directly associated with the feeding pattern of birds at feeding stations" not sure the strength of the hypothesis or whether I should reject it for an alternative such as "The time of day has no significance when birds feed at feeding stations"


I'd prefer the second as you are looking for an effect rather than looking for evidence that there is no effect (which is harder and would require a much larger sample size).
Reply 8
No excel was to be used for the graphs only and Minitab to be used to do the statistics - it's in minitab I am having difficulty working out what test to run for visits against time of day.
Reply 9
I am still pondering whether from the data I have would I reject my hypotheses "The time of day has no significance when birds feed at feeding stations" I am just learning about statistics and a little unsure, form the sum totals Morning 107 Afternoon 91 and Evening 41 I can see Morning and Afternoon are close and evening is significantly lower and therefore time of day does have significance on the feeding pattern of birds. Okay this is very simplistic and I cannot just draw my conclusion from one set of data. I am just struggling with the who concept of data Analysis, I want to improve but just cannot get my head around what type of test to use, how to break up my data so I can get the best graphs etc.

Quick Reply

Latest