Statistics - Hypothesis Testing Projects

ASSIGNMENT 3 PRACTICE DATA SETS There are many and varied forms of statistical analysis, suitable for

ASSIGNMENT 3

PRACTICE DATA SETS

There are many and varied forms of statistical analysis, suitable for a wide range of research designs. This unit of study will provide an introduction to a small, though popular, subset of statistical procedures. The techniques covered include those which are commonly found in basic randomised control trials.

To assist you in learning the principles of statistics and data analysis, and to use the statistical analysis software package, we have produced two data sets. The first is real. Although the second data set is computer-generated it is modelled on the results of a randomized controlled trial. It is intended that you will practice hand calculations with the Vukicevic data set. The other data set will be used for practice with a computer package.

HAND CALCULATION DATA SET

The hand calculation data set has been provided by Meri Vukicevic from La Trobe University. Her study is, as yet, unpublished. There were 48 participants in the study; 31 women and 17 men, with an average age of approximately 82 years. All were experiencing some vision loss, mostly maccular degeneration, which was interfering with their ability to read. The study’s purpose was to assess the effectiveness of two interventions designed to improve the number of words the participants could read in one minute.

One intervention was simply to provide participants with a large sheet of magnifying plastic which they could place over reading materials and which would enlarge the size of letter. The other intervention is called "eccentric viewing". It consists of training people to read using their peripheral vision. In all cases, peripheral vision was unaffected by the central visual loss.

Participants were randomly divided into four groups of 12. One group was a waiting-list control group, and received no intervention, but was assessed before and after the other groups were treated. One group received only the magnification treatment. One group received only the eccentric viewing treatment. One group received both magnification and eccentric viewing treatments. All participants had their reading speed measured before the treatment period, and again afterwards.

Magnification		Eccentric Viewing		Combination		Non-Intervention
Pre	Post	Pre	Post	Pre	Post	Pre	Post
93	123	35	42	72	147	0	0
115	195	213	216	126	165	77	72
102	98	21	52	0	47	86	90
72	112	0	52	132	233	107	103
210	282	26	73	123	222	63	60
18	192	70	66	189	256	117	120
216	252	73	88	48	136	141	195
216	240	57	114	44	108	12	15
24	58	16	156	88	177	98	70
63	95	35	101	5	73	81	95
25	102	43	125	88	165	56	42
35	125	107	188	111	201	35	38

COMPUTER PACKAGE DATA SET

The reference for the trial which will provide data for computer package practice is

Ward, J., & Sanson-Fisher, R. (1996). Does a 3-day workshop for family medicine improve preventive care? A randomized control trial. Preventive Medicine, 25, 741-747.

This study was conducted on a group of medical doctors attending postgraduate training workshops provided by the Royal Australian College of General Practitioners in the state of New South Wales. The workshops are three-day intensive sessions intended to provide further education to doctors undergoing postgraduate training. The study aimed to evaluate the effectiveness of such workshops in improving doctors' consultation habits in general practice in regard to questioning patients about their smoking status.

Doctors attending the postgraduate training were randomly assigned to one of two types of workshops. The 'control' workshops were on the topic of "rational prescribing". The 'experimental' workshops were on the topic of "disease prevention", particularly prevention through investigation of smoking status. It was predicted that doctors who attended the disease prevention workshops would subsequently ask more smoking-related questions of their patients than they did before the workshops, and more than the doctors in the control condition.

The principal measuring device was coding of audiotapes taken of the doctors' consultations with patients. The main coding recorded whether or not the doctor had asked her/his patients about their smoking status. Audiotaping of consultations was conducted both before and after the workshops. Consent for the taping was obtained from all parties. However, only the researchers were aware that the audiotapes were to be used to test the efficacy of the workshops.

Trained coders used a rating scale to code the recorded consultations. The coders did not know the identity of the doctors whose recordings they coded, or whether the

recording had been made before or after the workshops.

Thirty-four doctors, 15 female and 19 male, provided data for the study. Five females and 11 males participated in the "disease prevention" workshop which was the experimental intervention. Ten females and eight males participated in the "rational prescribing" workshop which was the control intervention. Together they performed clinical consultations with a total of 1362 patients who consented to participate in the study and whose audio recordings were codable. Six hundred and thirty nine consultations took place before the workshops and 723 after.

Two tables are produced below. Table 1 is a segment of Ward & Sanson-Fisher's Table 1 (1996, p 744). Table 2 is the (fictitious) results of the coding of consultations. The figures in Table 2 are, for each doctor, the percentage of consultations in which s/he inquired about the patients' smoking status.

TABLE 1
	Experimental Group N = 16 Trainees	Control Group N = 18 Trainees
Consultations with adults 16-65 years N audiotaped consultations Pre Post	314 325	346 377
N (%) consultations in which trainee asked about smoking status Pre Post	81 (26%) 97 (30%)	66 (19%) 57 (15%)

TABLE 2

Percentage of Consultations in which Trainees asked about Smoking Status

Experimental Pre-Workshop	Experimental Post-Workshop	Control Pre-Workshop	Control Post-Workshop
46	47	3	1
42	56	10	4
15	19	44	40
43	46	19	15
17	28	17	5
18	15	71	73
9	10	6	4
46	52	10	4
4	7	12	8
33	43	22	12
55	60	20	15
19	20	5	4
4	10	33	26
18	19	14	12
27	32	0	3
21	20	39	40
		8	3
		5	1

Ward & Sanson-Fisher helped conduct the training workshops whose effectiveness was evaluated in the study. They no doubt hoped the workshops would produce noticeable differences in the behavior of their doctors during consultations with patients.

ASSIGNMENT 3 Chi-Square, t-tests & ANOVA

PART 2 - Please use the Ward & Sanson-Fisher practice data set.
Task 1	Use analysis of variance to compare the Experimental and Control group scores before the workshop sessions.
Task 2	Write out in English and in symbol form a null hypothesis (other than that for Task 1) which can be tested using the information from the computer package data set.
Task 3	Using the computer package, do you accept or reject the null hypothesis in task 2? Explain why. (c) Explain in English what the analysis has told you.
Task 4	Write out a null hypothesis for a chi square test of independence which could be performed on the data.
Task 5	Perform the test of independence and explain what it tells you. Hint: If you are using SAS, use the SAS commands from Tasks 1 & 2 of Part 2, Assignment 2, then add SAS commands for chi square.