topic Re: How can i estimate the differences at different time? in Statistical Procedures

How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 07:48:55 GMT

Hi guys!

I came across a problem with a small sample, where we did measure the amount of fat in goat's milk during 2 year experiment (YEAR), where we took samples continuosly 5 times (SAMPLING) each year from 2 different groups (TYPE) of goats. And now we need to know if there was some kind of difference between each sampling in each year?

Can somebody help me with that?

Thank you,

Here is the data:

DATE	SAMPLING	YEAR	TYPE	FAT
14.07.2015	1	1	EKO	2,44
14.07.2015	1	1	KON	2,53
12.08.2015	2	1	EKO	3,06
12.08.2015	2	1	KON	2,84
9.09.2015	3	1	EKO	3,19
9.09.2015	3	1	KON	3,14
7.10.1015	4	1	EKO	2,84
7.10.2015	4	1	KON	3,00
4.11.2015	5	1	EKO	3,15
4.11.2015	5	1	KON	3,18
8.06.2016	1	2	EKO	2,70
8.06.2016	1	2	KON	2,86
6.07.2016	2	2	EKO	2,85
6.07.2016	2	2	KON	2,61
1.08.2016	3	2	EKO	2,75
1.08.2016	3	2	KON	2,73
7.09.2016	4	2	EKO	2,92
7.09.2016	4	2	KON	2,90
5.10.2016	5	2	EKO	3,29
5.10.2016	5	2	KON	3,61

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 09:02:51 GMT

Do a sort by type and date, the use the lag(FAT) function to compare values between observations.

You might also need the first. function to reset figures. You need to use set by; for this to function 🙂

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 09:22:09 GMT

Thank you for your help, but can i ask you to give me an example how to do it, step by step?

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 09:24:41 GMT

HI!

Can you supply a data step that creates the data?

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 09:27:00 GMT

Sorry i don't know what data do you mean. This data, which i paste it in first post is all i have.

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 09:36:42 GMT

Hi!

I ment the data that you posted, but I don't want to spend a lot of time just trying to create the test data set.

If you can supply that, it makes it a lot easier for me to help you 🙂

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 09:38:16 GMT

Oh ok, here it is.

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 09:45:48 GMT

Thanks!

Is it the difference between samplings, i.e: between 1 and 2, 2 and 3, 3 and 4 or between sampling and year, i.e: sampling 1 year 1 and sampling 1 year2, sampling 2 year 1 and sampling 2 year2....?

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 09:49:06 GMT

It should be, i would like to estimate if there is any difference between two continuos sampling in within a year and type.

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 10:01:15 GMT

proc sort data=fat;
 by type year date;
run;

data want;
 set fat;
 by type year date;

 prev_fat=lag(fat);
 diff_fat=dif(fat);

 if first.year then do;
	prev_fat = .;
	diff_fat = .;
 end;

run;

Will this work?

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 10:07:57 GMT

I'm really new to those kind of procedures, so i must fill anything with before i run it?

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 10:08:54 GMT

I sorted the old data, but now i can't make a new (want) data.

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 10:13:33 GMT

Ok, i made it, but is it also possible to get if those differences are statistically different?

Re: How can i estimate the differences at different time?

FredrikE — Mon, 05 Feb 2018 10:49:23 GMT

Sorry, but I think that is out of my knowledge 🙂

//Fredrik

Re: How can i estimate the differences at different time?

BorutK — Mon, 05 Feb 2018 10:51:52 GMT

Thank you anyway, i learned couple of new things from you anyway 🙂 I tried with proc mixed and then with lsmeans but i don't think that is correct. It is really hard to work with so small number of samples.

Re: How can i estimate the differences at different time?

Reeza — Mon, 05 Feb 2018 15:59:05 GMT

Here is what @FredrikE was referring to when he was asking for a data step:

https://communities.sas.com/t5/SAS-Communities-Library/How-to-create-a-data-step-version-of-your-data-AKA-generate/ta-p/258712

I suspect you're looking for a statstical analysis rather than just a calculation of the previous difference.

First for an analysis you need to determine the type of analysis required, ie ANOVA, MANOVA, etc.

Then you can program it. It sounds like you're still in the trying to determine how you should analyze your data.

This question is less likely to be answered here since it's methodological question, rather than pure programming. You may get better responses on Cross Validated.

@BorutK wrote:

Hi guys!

I came across a problem with a small sample, where we did measure the amount of fat in goat's milk during 2 year experiment (YEAR), where we took samples continuosly 5 times (SAMPLING) each year from 2 different groups (TYPE) of goats. And now we need to know if there was some kind of difference between each sampling in each year?

Can somebody help me with that?

Thank you,

Here is the data:

DATE SAMPLING YEAR TYPE FAT

14.07.2015 1 1 EKO 2,44

14.07.2015 1 1 KON 2,53

12.08.2015 2 1 EKO 3,06

12.08.2015 2 1 KON 2,84

9.09.2015 3 1 EKO 3,19

9.09.2015 3 1 KON 3,14

7.10.1015 4 1 EKO 2,84

7.10.2015 4 1 KON 3,00

4.11.2015 5 1 EKO 3,15

4.11.2015 5 1 KON 3,18

8.06.2016 1 2 EKO 2,70

8.06.2016 1 2 KON 2,86

6.07.2016 2 2 EKO 2,85

6.07.2016 2 2 KON 2,61

1.08.2016 3 2 EKO 2,75

1.08.2016 3 2 KON 2,73

7.09.2016 4 2 EKO 2,92

7.09.2016 4 2 KON 2,90

5.10.2016 5 2 EKO 3,29

5.10.2016 5 2 KON 3,61

Re: How can i estimate the differences at different time?

ballardw — Mon, 05 Feb 2018 17:25:28 GMT

You say "2 different groups (TYPE) of goats". Was there only one goat of each type? And what would the FAT variable represent? If it is the average of the goats in the group and there were more than one goat per type you might consider using the individual goat measurements instead of an average or group summary. That might help the sample size a bit.

Re: How can i estimate the differences at different time?

BorutK — Tue, 06 Feb 2018 06:58:26 GMT

We took a sample from each group of goats, so yes it the average of the group. I agree that it would be much easier if we would have individual measurements, but only have this.

Re: How can i estimate the differences at different time?

ballardw — Tue, 06 Feb 2018 16:07:17 GMT

You might be able to include the number of goats in each sample group and indicate that value as a frequency variable, often with the FREQ statement. With a mean value that would have the procedure treat the data as N identical measures, changing the sample size though confidence limits could be somewhat restricted.