turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- How to compare two rates in SAS - Need Help, Pleas...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 10:52 AM

I know this may be a simple answer, but I cannot seem to find what I am looking for via my internet searches and thought I would ask here.

I have two datasets. Each dataset has an up to date vaccine variable. There is one dataset for year 2010 and one dataset for year 2011.

I am wanting to see if the rate of individuals that were up to date in year 2010 is significantly different compared to the rate of the individuals up to date in year 2011.

Year 2011 UTD Variable

0 30

1 80

37.5% were up to date

Year 2010 UTD Variable

0 50

1 75

66.67% were up to date

I was able to test if the difference between those were up to date and those who were not was statistically significant by using proc freq and the chisq option.

Just not sure how to compare rates from two different datasets.

Thank you for any help you can provide.

Accepted Solutions

Solution

10-12-2011
09:54 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 09:54 PM

Not sure what test you want. Here is how to do a CHISQ test.

data have ;

input year utd count;

cards;

2011 0 30

2011 1 80

2010 0 50

2010 1 75

run;

proc freq ;

weight count;

tables year*utd / chisq;

run;

The FREQ Procedure

Table of year by utd

year utd

Frequency|

Percent |

Row Pct |

Col Pct | 0| 1| Total

---------+--------+--------+

2010 | 50 | 75 | 125

| 21.28 | 31.91 | 53.19

| 40.00 | 60.00 |

| 62.50 | 48.39 |

---------+--------+--------+

2011 | 30 | 80 | 110

| 12.77 | 34.04 | 46.81

| 27.27 | 72.73 |

| 37.50 | 51.61 |

---------+--------+--------+

Total 80 155 235

34.04 65.96 100.00

Statistics for Table of year by utd

Statistic DF Value Prob

------------------------------------------------------

Chi-Square 1 4.2210 0.0399

Likelihood Ratio Chi-Square 1 4.2567 0.0391

Continuity Adj. Chi-Square 1 3.6732 0.0553

Mantel-Haenszel Chi-Square 1 4.2031 0.0404

Phi Coefficient 0.1340

Contingency Coefficient 0.1328

Cramer's V 0.1340

Fisher's Exact Test

----------------------------------

Cell (1,1) Frequency (F) 50

Left-sided Pr <= F 0.9861

Right-sided Pr >= F 0.0273

Table Probability (P) 0.0134

Two-sided Pr <= P 0.0531

Sample Size = 235

All Replies

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 07:49 PM

As far as I know you can't.

You'll need to get them into a single dataset...there's a ton of ways to do this but one way is to set the data together and include the year, then run proc freq.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 09:04 PM

How about PROC COMPARE? This assumes you have one row in each table for each individual and you have a common key identifying the individuals. You can also configure the method used to identify differences and the size of the difference.

**proc** **compare** base = dataset2010

compare = dataset2011

out = difs

OUTNOEQUAL LISTEQUALVAR LISTCOMPVAR LISTBASEVAR

MAXPRINT=**300**

;

id individual_ID;

var rate;

where UTD = 1;

**run**;

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-18-2011 02:44 PM

I will have to try this method and see how it works. I have not yet been able to try it on my data but will let you know if it works out. thank you for the suggestion!

Solution

10-12-2011
09:54 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 09:54 PM

Not sure what test you want. Here is how to do a CHISQ test.

data have ;

input year utd count;

cards;

2011 0 30

2011 1 80

2010 0 50

2010 1 75

run;

proc freq ;

weight count;

tables year*utd / chisq;

run;

The FREQ Procedure

Table of year by utd

year utd

Frequency|

Percent |

Row Pct |

Col Pct | 0| 1| Total

---------+--------+--------+

2010 | 50 | 75 | 125

| 21.28 | 31.91 | 53.19

| 40.00 | 60.00 |

| 62.50 | 48.39 |

---------+--------+--------+

2011 | 30 | 80 | 110

| 12.77 | 34.04 | 46.81

| 27.27 | 72.73 |

| 37.50 | 51.61 |

---------+--------+--------+

Total 80 155 235

34.04 65.96 100.00

Statistics for Table of year by utd

Statistic DF Value Prob

------------------------------------------------------

Chi-Square 1 4.2210 0.0399

Likelihood Ratio Chi-Square 1 4.2567 0.0391

Continuity Adj. Chi-Square 1 3.6732 0.0553

Mantel-Haenszel Chi-Square 1 4.2031 0.0404

Phi Coefficient 0.1340

Contingency Coefficient 0.1328

Cramer's V 0.1340

Fisher's Exact Test

----------------------------------

Cell (1,1) Frequency (F) 50

Left-sided Pr <= F 0.9861

Right-sided Pr >= F 0.0273

Table Probability (P) 0.0134

Two-sided Pr <= P 0.0531

Sample Size = 235

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-13-2011 02:03 AM

There is a problem. That is correlation.

If these two year's experiments were applied at the same patient. then there is a correlated effect.

You can not directly use these data into proc freq .

Need to subtract between them to remove this correlated effect.

Ksharp

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-18-2011 02:42 PM

Thank you pointing this out Ksharp. I need to keep this in mind and take into account the correlated effect.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

10-12-2011 11:42 PM

I've never tried this, but just found it in a 2007 SAS-L thread. I think you should just get the numbers from running the two proc freqs and then apply them to a "proportions test".

Here is what I discovered in that 2007 thread:

Try the SAS built-in tool for proportion test. It's under

Solutions-->Analysis-->Analyst to open the Analyst Window; then

Statistics-->Hypothesis Tests--> One-Sample (or Two-Sample) Test for

Proportions. You can only test the equality of proportions between two

regions each time (I am not sure if this is true or I haven't just

found the right option)