Dilemma: Conducting Statistical Tests Using Proc Surveyfreq Statement on SAS 9.4

[yli33]

Hi Everyone,

I have attempted to conduct chi-square tests using my research group's survey dataset on SAS 9.4; however, SAS was unable to compute the required chi-square statistics/p-values using my dataset due to the following reason:

NOTE: Chi-square tests cannot be computed for the table of SEX by LANG because at least one table

cell has 0 frequency.

Frankly, this error message is highly problematic because most of the dataset contain cells with 0 frequency after conducting a cross-tabulation of variables on SAS. Therefore, I would like to inquire whether any expert in the community may know of a different type of statistical tests that are analogous to the wald chi-square test in terms of determining the relationship between the independent variable and dependent variable by computing p-values. Thank you very much in advance!

Best,

Lisa

[PaigeMiller]

It's not the error message that's problematic, it is your data that is problematic.

Can you share with us the number of data points in each cell, and the code you are using? 

[yli33]

Hi Paige,

Thank you for your response. Please find below the sample code for one of the cross-tabulations.

proc surveyfreq data=wdp.newzhang;

tables sex*lang / wchisq;

weight weight_3;

run;

As for the number of data points in each cell, I am not exactly sure what you specifically mean, such as the weighted frequencies for each cell level? Thanks so much!

[PaigeMiller]

Yes, the weighted frequency in each cell is what I want. 

[yli33]

Hi Paige,

Please find below a sample dataset that includes the percentages and weighted frequencies. Thanks!

Immigration Status (%, n)		Total=1428	Male=1391	Female=37.10
	Born in the U.S.	30.66 (437.84)	29.54 (421.91)	1.12 (15.93)
	Naturalized Citizen	4.78 (68.24)	4.42 (63.08)	0.36 (5.16)
	Legal Resident	16.10 (230.01)	15.82 (225.94)	0.28 (4.08)
	Visa Worker	5.06 (72.40)	4.93 (70.46)	0.14 (1.94)
	Undocumented	31.83 (454.62)	31.42 (448.69)	0.41 (5.92)
	Refuse	11.55 (164.98)	11.26 (160.91)	0.28 (4.08)

[yli33]

Hi Paige,

Please ignore the previous dataset because I just realized that I was able to compute a wchisq p-value using that set of data.

Please find the new dataset below that is problematic.

Table of WATER by HEAT_1 WATER HEAT_1 Frequency Weighted
Frequency Std Err of
Wgt Freq Percent Std Err of
Percent EMPLOYER 0   1   2   3   4   Total GC 0   1   2   3   4   Total NONE 0   1   2   3   4   Total REFUSE 0   1   2   3   4   Total SELF 0   1   2   3   4   Total SUB 0   1   2   3   4   Total Total 0   1   2   3   4   Total Frequency Missing = 1

3	4.48080	3.57479	0.3141	0.2502
186	170.80100	17.33165	11.9719	1.1794
201	215.70530	20.02574	15.1193	1.3390
226	203.39680	18.87854	14.2566	1.2752
138	188.13820	19.72814	13.1871	1.3156
754	782.52210	33.74747	54.8489	1.8157
1	1.67200	1.67200	0.1172	0.1172
41	26.80870	7.21888	1.8791	0.5035
38	35.50600	10.77292	2.4887	0.7445
33	18.61870	4.62034	1.3050	0.3249
27	31.25230	7.64386	2.1906	0.5330
140	113.85770	15.62050	7.9806	1.0632
0	.	.	.	.
4	4.77570	2.57006	0.3347	0.1802
2	5.16230	3.86906	0.3618	0.2707
2	1.56880	1.29023	0.1100	0.0905
1	3.49030	3.49030	0.2446	0.2443
9	14.99710	5.94218	1.0512	0.4148
0	.	.	.	.
0	.	.	.	.
2	0.88580	0.63785	0.0621	0.0448
0	.	.	.	.
1	1.67200	1.67200	0.1172	0.1172
3	2.55780	1.78896	0.1793	0.1254
3	5.53350	3.88648	0.3879	0.2719
105	102.17350	13.10830	7.1616	0.9071
117	116.41740	14.10271	8.1600	0.9726
165	140.71050	14.41749	9.8627	1.0025
122	129.44480	14.39659	9.0731	0.9956
512	494.27970	25.93700	34.6453	1.6986
0	.	.	.	.
2	1.41380	1.26109	0.0991	0.0884
6	6.83570	3.04878	0.4791	0.2138
6	5.52040	2.62661	0.3869	0.1842
2	4.70230	4.03727	0.3296	0.2824
16	18.47220	5.82380	1.2948	0.4071
7	11.68630	5.53366	0.8191	0.3864
338	305.97270	22.25651	21.4464	1.4803
366	380.51250	26.12400	26.6711	1.6589
432	369.81520	23.27418	25.9213	1.5525
291	358.69990	25.16904	25.1422	1.6174
1434	1427	36.26149	100.000

[yli33]

Hi Paige,

Please again ignore the dataset that I just submitted as the formatting is entirely distorted.

Please use the dataset below. Thanks!

Language Fluency (%, n)		Total=1428	Male=1391	Female=37.10
	English	18.64 (266.26)	18.14 (259.10)	0.50 (7.16)
	Spanish	50.54 (721.88)	49.67 (709.36)	0.88 (12.52)
	Both	28.80 (411.43)	27.75 (396.30)	1.06 (15.14)
	Other	0.46 (6.64)	0.46 (6.64)	6.15 (2.28)

[PaigeMiller]

I find this impossible to understand. 

[yli33]

Hi Paige,

Just to clarify, the weighted frequencies are all within the parentheses following the percentages. Hope this helps a bit! Thanks!

[PaigeMiller]

This table gives the exact same error message quoted above, that there is a cell with zero frequency?

[yli33]

Hi Paige,

I think the attached document with the original cross-tabulation table with sex by language should provide a more clear overview of the problem. The previous dataset that I have posted was the modified table after combining the weighted frequencies of several groups.

[PaigeMiller]

I don't download Microsoft Office documents, as they can be security threats. Please place your information in a text file.

[yli33]

Hi Paige,

I have converted the word document into a text document. Thanks!

Best,

Lisa

[PaigeMiller]

Ok, this isn't working.

I asked a simple yes or no question. I don't see where you have answered the question. I asked:

This table gives the exact same error message quoted above, that there is a cell with zero frequency?

[yli33]

Hi Paige,

I apologize for missing your questions earlier. Nevertheless, the table with Sex by Lang does show the error message that at least one table cell has 0 frequency. Thanks!