Semi-Structured Interviews in This Study. The Subjects

¶ … semi-structured interviews in this study. The subjects were randomly divided into two experimental groups and one control group. The treatment and control groups were as follows: (A) Family therapy plus Social Skills Training (SST) plus a placebo (B) Administration of Fluoxetine; (C) Family therapy plus Social Skills Training (SST) (Control Group). All subjects were assessed and treated for substance abuse problems regardless of group assignment. A total of 9 subjects were included in the study. All treatment took place in clinical settings and was configured to be individual rather than group treatment.

Quantitative data was obtained from the closed-end items of the semi-structured interviews. That is, some parts of the interview were structured like a survey with responses indicated by marking answers on a 5-point Likert scale. The choices in the Likert scale were as follows: Strongly Agree = 5; Agree = 4; Neutral or Not Sure = 3; Disagree = 2; Strongly Disagree = 1. The questions were focused on the subjects' perceptions of their use of controlled substances before and after drug and alcohol treatment, and are listed below.

Q1

Self-medicating with alcohol or marijuana is not a good idea.

Q2

You like your life better when you drink regularly with friends.

Q3

Your friends like you better when you smoke marijuana with them.

Q4

People think you are a jerk if you drink too much.

Q5

Alcoholics don't get what they want out of life.

Q6

I have more choices when I don't do drugs.

(a) Statistical Analysis Choices.

Standard Error Bars. Using standard error bars, the means of different populations can be compared where the independent variable is qualitative and the dependent variable is quantitative -- in this study, the scores assigned to item questions by subjects (Radford, 2011). In this study, the coded narrative response on the semi-structured portion of the interview places subjects in one of three particular categories: (1) No change in perceptions; (2) moderately changed perceptions, or (3) substantively changed perceptions (Radford, 2011).

Chi-square. Chi-square uses contingency tables to analyze the relationships between the subjects' scores for two categorical or qualitative variables (Lane, 2006). Chi-square allows the researcher to determine if the hypothesis has been supported or not supported (Lane, 2006). In the example below, the cells contain frequencies of responses that have been weighted according to the following schema, and which is equivalent to the "Top Two Box" format used to report responses on survey instruments (Lane, 2006). Note that the weighting for the two Question Sets shows an inverse relationship.

For Question Set 1 -- 3.

Strongly Agree = 5 and Agree = 4 on the Likert scale are interpreted as Substantive Change. Disagree = 2; Strongly Disagree = 1 on the Likert scale are interpreted as No Change. Neutral or Not Sure = 3 on the Likert scale is interpreted as Moderate Change.

For Question Set 4 -- 6.

Disagree = 2; Strongly Disagree = 1 on the Likert scale are interpreted as Substantive Change. Strongly Agree = 5 and Agree = 4 on the Likert scale are interpreted as No Change. Neutral or Not Sure = 3 on the Likert scale is interpreted as Moderate Change

Question Set 1-3

No Change

Moderate Change

Substantive Change

Treatment Group A

2

1

6

9

Treatment Group B

2

4

3

9

Control Group C

0

3

6

9

Total*

4

8

15

27

*Numbers are simulations

The data in the contingency chart above show that subjects receiving family therapy plus social skills training plus a placebo showed greater shifts in their perceptions about substance abuse than did the subjects who received only pharmacological interventions. Interestingly, the subjects in the control group also showed strong shifts in their perceptions about substance abuse.

(b) Justification of Statistics Used.

Standard error bars are an approximation of a t-test, but with some advantages (Radford, 2011). Standard error bars provide a visual way to evaluate statistical differences by graphically illustrating the variability of the data (Radford, 2011). Also, standard error bars can be used to compare more than two means, which is useful in this analysis because there is a mean score for the positively rated question set (Question 1 -- 3) and a mean score for the negatively rated question set (Questions 4-6 ) (Radford, 2011). The main caveat is that there is a greater threat of obtaining false positive result -- that is, the two populations are thought to be different when there is no real difference (Radford, 2011). A prudent follow-up to standard error bars, then, is to perform an actual t-test or an ANOVA (Radford, 2011).

(c) Results Section.

Three of the items (Questions 1 -- 3) are scored as positive responses, such that the points on the Likert scale are considered additive with accumulated points representing the subjects' ability to address their substance abuse problem. Three of the items (Questions 4 -6) are scored as negative responses, such that the points on the Likert scale indicate maladaptive perceptions of the subjects with regard to their ability to address their substance abuse problems.