topic Re: Different results from different import data path in SAS Procedures

Different results from different import data path

Golf — Wed, 30 Sep 2020 00:58:42 GMT

Hello,

When the same data set are imported from different data path, I found that the result from "proc arima" give slightly different results. Have anyone found this problem?

Thanks.

Re: Different results from different import data path

Reeza — Wed, 30 Sep 2020 01:34:00 GMT

What do you mean by "imported from different data path"?

If you suspect your imports are giving you different data for whatever reason, use each method and then use PROC COMPARE on the two data sets to see if they're different.

proc compare data=have1 compare=have2;
run;

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:26:45 GMT

Hi Reeza,

Here is my meaning of the import data path.

My student downloaded the excel file from my google classroom into her data path: "/home/u49195926/SM518/ch3_5.xlsx". Using the following procedure :

The results from using student's data path: "/home/u49195926/SM518/ch3_5.xlsx" are as follows:

But when using my data path"/home/u45083887/my_content/SM518/SM518 Ch3/ch3_5.xlsx", the results are different as shown below. (We make sure that the data in excel are the same for mine and students)

I really suspect how this could be occurred.

Thank You.

Re: Different results from different import data path

Reeza — Wed, 30 Sep 2020 14:31:01 GMT

No results shown but my guess is you're likely not using the same versions of SAS most likely then.
Check your versions using PROC PRODUCT_STATUS;RUN: and see what version of SAS/STAT you're each running.

Re: Different results from different import data path

ballardw — Wed, 30 Sep 2020 15:16:21 GMT

You don't show your output or how it differs.

If the differences are small in calculated items like BIC, say in the 5th decimal place only or possibly the 4th and 5th then different versions of SAS where algorithms get modified between releases is a very a likely possibility.

@Golf wrote:

Hi Reeza,

Here is my meaning of the import data path.

My student downloaded the excel file from my google classroom into her data path: "/home/u49195926/SM518/ch3_5.xlsx". Using the following procedure :

The results from using student's data path: "/home/u49195926/SM518/ch3_5.xlsx" are as follows:

But when using my data path"/home/u45083887/my_content/SM518/SM518 Ch3/ch3_5.xlsx", the results are different as shown below. (We make sure that the data in excel are the same for mine and students)

I really suspect how this could be occurred.

Thank You.

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:15:00 GMT

Thank You

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:19:12 GMT

Thank You.

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:28:34 GMT

I don't know why my results did not show up in the post.

Here is my results, they are different.

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:39:37 GMT

I just check the version of mine and a student. They are the same.

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 01:50:26 GMT

I just check the version of mine and a student. They are the same.

Re: Different results from different import data path

Reeza — Thu, 01 Oct 2020 02:49:04 GMT

What version and can you post the data so we can check the results?

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 05:19:06 GMT

Hi Reeza,

Here is my version. (The data is after the version detail)

For Base SAS Software ...

Custom version information: 9.4_M6

Image version information: 9.04.01M6P110718

For SAS/STAT ...

Custom version information: 15.1

For SAS/GRAPH ...

Custom version information: 9.4_M6

For SAS/ETS ...

Custom version information: 15.1

For SAS/OR ...

Custom version information: 15.1

Image version information: 9.04.01M6P050819

For SAS/IML ...

Custom version information: 15.1

For SAS/QC ...

Custom version information: 15.1

For SAS/CONNECT ...

Custom version information: 9.4_M6

For SAS Enterprise Miner ...

Custom version information: 15.1

For SAS Time Series Workspace Macros ...

Custom version information: 15.1

Image version information: 9.04.01M5P110718

For SAS/ACCESS to Postgres ...

Custom version information: 9.4_M6

For SAS Integration Technologies ...

Custom version information: 9.4_M6

For SAS/Secure 168-bit ...

Custom version information: 9.41_M3

For SAS Credit Scoring ...

Custom version information: 15.1

For SAS Text Miner ...

Custom version information: 15.1

For SAS High-Performance Forecasting ...

Custom version information: 15.1

For High Performance Suite ...

Custom version information: 2.2_M7

For SAS Forecast Server Mid-Tier ...

Custom version information: 15.1

Image version information: 9.04.01M5P110718

For SAS/ACCESS Interface to PC Files ...

Custom version information: 9.4_M6

For SAS/ACCESS Interface to MySQL ...

Custom version information: 9.4_M6

Here is the data.

Export
9.7323
9.9386
10.5581
9.5671
10.2601
10.1751
10.3672
10.0169
10.4475
10.5712
10.9055
10.3344
10.9132
11.0472
11.0752
10.8949
11.1713
11.3635
11.5535
11.1439
11.5798
11.9851
12.0605
11.5752
11.938
12.2423
12.232
11.861
12.1799
12.754
12.6888
12.2581
12.4873
13.0316
12.9496
12.5721
13.0557
13.3554
13.4013
13.0221
13.3251
13.764
13.6152
13.4049
13.6034
14.1636
14.0264
13.7761
14.0608
14.5765
14.5491
14.3401
14.5873
15.1227
14.966
14.8268
15.2012
15.6293
15.6306
15.1636
15.6866
16.1653
15.8831
15.7871
15.9769
16.6352
16.3872
16.2214
16.5239
17.0421
16.9188
16.7274
17.0048
17.5631
17.4562
17.245
17.4456
18.1318
17.996
17.7068
17.9141
18.3945
18.3056
18.1098
18.3128
19.1285
18.8676
18.5391
18.7751
19.4459
19.2726
18.9861
19.1439
20.003
19.6391
19.6406
19.8099
20.4684
20.3344
19.799
20.2407
21.0073
20.7403
20.3713
20.8789
21.3626
21.1808
20.9169
21.1793
21.9515
21.6688
21.3734
21.6468
22.36
21.9971
21.6864
22.1199
22.7439
22.4872
22.2034
22.5794
23.1908
22.8497
22.5202
22.9665
23.635
23.3512
23.0428
23.3996
24.0955
23.7421
23.4174
23.6984
24.5057
24.1983
23.7094
24.2367
24.9358
24.5153
24.0533
24.5504
25.1616
25.0315
24.467
24.9887
25.6824
25.4345
25.1052
25.3507
26.0501

Re: Different results from different import data path

Kurt_Bremser — Thu, 01 Oct 2020 07:23:05 GMT

Get rid of the problem points.

Save your data as a csv file; create a data step to read that. Hand your students that csv file and the data step, so you can make sure that the data is read in the exactly same way; IMPORT of Excel files is subject to many externalities (like locale) and makes guesses, which means that it is always unreliable.

Run a checksum (md5) on your csv file and keep it for reference.

Make sure that downloads/uploads of files are done consistently (same software, same settings).

In the current situation, have the student download the .sas7bdat file of the dataset as is, upload it to your environment, and run a PROC COMPARE with your dataset.

Re: Different results from different import data path

Kurt_Bremser — Thu, 01 Oct 2020 07:28:02 GMT

PS if you still get differences when using CSV as transport medium, post the CSV file by copy/pasting it into a code box:

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 11:16:46 GMT

Thank You

Re: Different results from different import data path

Golf — Thu, 01 Oct 2020 12:41:19 GMT

Hello KurtBremser,

I save the excel file into CSV and let student download it from google classroom. The results totally changed from the previous outputs. I attached the CSV file.

Thank You.

PS I notice that if I sent the excel file directly to student via LINE application, the results will be exactly same. When students download excel from google classroom, the results different.

Re: Different results from different import data path

Reeza — Thu, 01 Oct 2020 15:22:56 GMT

Have you checked that the Excel File isn't being changed somehow? Have a student read in the data set on their system, send you the dataset as a SAS dataset (sas7bdat file) and then run the compare on it.

You also have nothing controlling the order of these observations, is it possible someone sorted it different?

I'm getting the same results as your results below here.

You can also get them to run PROC MEANS on the data set and compare your outputs to ensure you have the exact same data.

@Golf wrote:

Hello KurtBremser,

I save the excel file into CSV and let student download it from google classroom. The results totally changed from the previous outputs. I attached the CSV file.

Thank You.

PS I notice that if I sent the excel file directly to student via LINE application, the results will be exactly same. When students download excel from google classroom, the results different.