namf
Forum Elite
Topics: 80
Posts: 312
|
HI all,
I guess stats qs belong here in the Behavioral science board?
Don't know how to handle questions whenever I see the following:
1)ANOVA
2)chi square
3)t-test (either matched pair or not)
4)Pearson correlation coefficient
Anyone want to teach? Thank you!
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
This info culled from FA and BRS Behav.Sci.
t-test: difference between the means of 2 samples.
-independent (nonpaired) t-test: two groups of subjects are sampled at the same time.
-dependent (paired) t-test: the same people from same group are sampled at two separate times.
ANOVA (ANalysis of VAriance): differences between the means of 3 or more samples.
Chi-Square (X^2): differences between frequencies in a sample, comparing percentages or proportions in 2 or more groups.
Pearson correlation coefficient (r): degree of relationship between 2 continuous variables with r always ranging between -1 and +1. If the two variables move in the same direction, r is positive (e.g., as calorie intake decreases, body weight decreases). If the two variables move in the opposite direction of each other, r is negative (e.g., as exercising increases, body weight decreases).
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
Here's some example questions adapted from BRS Beh.Sci. to try out:
1. Which statistical test is most appropriate for evaluating differences in the percentage of women who lose weight on low-carb diet versus the percentage of women who lose weight on protein-rich diet?
A) Paired t-test
B) Independent t-test
C) ANOVA
D) Chi-square test
E) Correlation
2. Which statistical test best differentiates initial body weight and final body weight for women on a low-carb diet?
A) Paired t-test
B) Independent t-test
C) ANOVA
D) Chi-square test
E) Correlation
3. Which test evaluates the relationship between body weight and systolic blood pressure in a group of 25-year-old women?
A) Paired t-test
B) Independent t-test
C) ANOVA
D) Chi-square test
E) Correlation
|
| mjl1717
Forum Hero
Topics: 955
Posts: 5,450
|
1)Chi
2)ANOVA
3)Correlation
PsychDr2b-Either way could you explain #2
ThankYou
___________________
Smell the coffee! "Is That an Osler move??"
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
Here's the answers with explanation:
1-D. Chi-square used to examine differencies between frequencies in a sample; in this question, the percentages of women who lose weight on low-carb diets versus the percentages of women who lose weight on protein-rich diets.
2-A. The t-test is used to examine the differences between the means of two samples. This is an example of a dependent (PAIRED) t-test becaus ethe same women are examined on two different occasions (Time 1: initial body weight before diet. Time 2: final body weight after diet).
3-E. Correlation is used to examine the relationship between two consecutive variables; in this question, systolic blood pressure and body weight.
|
| namf
Forum Elite
Topics: 80
Posts: 312
|
Thank you for going the extra-mile, PsychDr2B--this is really outstanding! Now lets see if I understand it:
1)Choice A, paired t-test, was my initial thought since we are comparing the amount of weight lost in the protein-eaters compared to the low-carb group ("t-test: difference between the means of 2 samples"). On further thought, the question asks for the percentage of women who lost weight in both groups--so using the mean weight lost in each group and paired t-test analysis appeared incorrect. Since chi-square was defined as "the differences between frequencies in a sample, comparing percentages or proportions in 2 or more groups"--I decided to choose D.
FINAL ANSWER: D
2)Not really sure about this one. Based on the definition "dependent (paired) t-test: the same people from same group are sampled at two separate times", there is applicability to the women's before-and-after weights on the low-carb diet. However, the t-test was defined as the "difference between the means of 2 samples"--and here we are looking at the individual weights, not the group means.
FINAL ANSWER:A
3)This appears to be a Pearson correlation coefficient based on the definition that it is "degree of relationship between 2 continuous variables"--bp and body weight in this example.
FINAL ANSWER: E
After all, I have to admit I don't feel like I really have the hang of this stuff--especially the chi-square vs. t-test difference. If you feel up to posting anymore questions, I would be greatful.
Thank you very much, PsychDr2B!
_______________________________________________
Fun is 99% inspiration and 1% perspiration (Take that!)
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
Good answers. Just to point out before we proceed, according to Kaplan notes, the USMLE does NOT require us to compute any of these statistical tests, but only to recognize what they are and when they should be used. I already summarized above the definitions as best as I could (which I think are listed and explained better in BRS Beh.Sci. than in Kaplan). So onto this next question (I'll post the explanantion, which helps more, after some of you try to answer it first):
A pharmaceutical researcher is examining the ulcerogenic potential of a new nonsteroidal anti-inflammatory drug. He gives 20 rats a single subcutaneous injection of the drug every day for one week, and gives a similarly matched group of 20 animals daily saline injections for one week. 24 hours after the final injection, the investigator sacrifices the rats, removes their stomachs, and examines them to determine if any ulcers were produced. He obtains the following data:
__________Ulcers present______Ulcers absent
Drug:_________12____________________8
No Drug:________8___________________12
Which of the following tests would be most appropriate for determining if administration of the drug increased the
incidence of stomach ulcers?
A. Analysis of variance (ANOVA)
B. Chi-squared test
C. Linear regression
D. Paired t-test
E. Pearson correlation coefficient
|
| rida
Forum Guru
Topics: 109
Posts: 721
|
B) Chi squared test
___________________
"If He takes you to it, He'll take you through it."
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
Good job Rida
Explanation of above question:
The correct answer is B. The most commonly used method for calculating p values from a two-by-two
contingency table is the chi-squared test. This test is used for frequency data (such as those above) rather
than for comparison of means. Since the investigator scored the stomachs as either containing ulcers or not
containing ulcers, without attempting to quantify the number of ulcers, the data represent frequencies rather
than means. Chi-squared is calculated as the sum for all cells of (Observed - Expected)2/Expected. The p value
for this value of chi-squared is obtained from a table using degrees of freedom equal to (number of rows - 1) x
(number of columns - 1).
ANOVA (choice A) is used to determine if the difference between two or more groups is significantly different.
This method is not applicable to raw frequency data.
Linear regression (choice C) is the process of fitting a straight line to a set of correlational data points by
minimizing the sum of squares of the vertical distances from the points to the line. This method would not be
suitable for the above data.
A paired t-test (choice D) is used to compare the means of two groups. It is not used to compare raw frequency
data.
Correlation is used when two variables are simultaneously measured in a sample. The Pearson correlation
coefficient (choice E) is used when both the independent and dependent variable are continuous. In this case,
the independent variable is not continuous (since the drug is either given or not given), and the dependent
variable is also not continuous (the ulcers are either present or absent).
Hope it helps.
|
| PsychDr2B
Forum Elite
Topics: 35
Posts: 197
|
This will probably be my last post on the subject, if you need anymore questions on this please PM me. Thanks
|
| namf
Forum Elite
Topics: 80
Posts: 312
|
T
|
| namf
Forum Elite
Topics: 80
Posts: 312
|
Thanks for all your help, PsychDr2B. Would I be making a miserably wrong guess if I thought that you enjoy research--hence the reason for the interest in the stats??
Whatever the reason--it is helpful: thank you 
|
|
| |
| | | | | | | | | | | | |
