I need help

#1
Hi guys,

I'm a Physiotherapy student in the process of writing my dissertation. I've done data collection for my project but now I'm at the point where I need to analyse said data and this is where I am getting stuck.
The aim of the study is to measure pain threshold following the application of treatment A or B. I had two groups of equal numbers, measured pain threshold for each participant but only treated one leg and left the other as control. Should I go for 2 sample t test or one way anova using my 'control' leg as a 3rd group?
 

Miner

TS Contributor
#2
I am in a different field, so I do not know the common protocols in your field. I would probably approach this differently, using the difference between the treatment leg and the control leg as the response, then comparing that between the two different treatments.

What is your data type? Continuous? Or an ordinal rating? That would affect your choice of parametric or nonparametric tests.
 
#3
I am in a different field, so I do not know the common protocols in your field. I would probably approach this differently, using the difference between the treatment leg and the control leg as the response, then comparing that between the two different treatments.

What is your data type? Continuous? Or an ordinal rating? That would affect your choice of parametric or nonparametric tests.
Well this is how useless I am. I don't know what type my data is. My data is basically a numerical value expressed in newtons which tells me how much pressure was applied to the leg.
 

Miner

TS Contributor
#5
That would be continuous data. You can use a 2-sample t-test on the difference between Delta A and Delta B. While this is not my field, I do know that different people have different pain thresholds. Using the delta for each person will remove this source of variation. Another thing to consider would be to include a placebo as a third treatment to adjust for the possibility of a placebo effect. If you include this, use a 1-way ANOVA to analyze the deltas, followed by Dunnett's test to compare treatments A and B to the control (placebo).

1612363920817.png
 
Last edited:
#6
That would be continuous data. You can use a 2-sample t-test on the difference between Delta A and Delta B. While this is not my field, I do know that different people have different pain thresholds. Using the delta for each person will remove this source of variation. Another thing to consider would be to include a placebo as a third treatment to adjust for the possibility of a placebo effect. If you include this, use a 1-way ANOVA to analyze the deltas.

View attachment 3042
Thanks for your reply. Data collection is over now so I can't use a placebo as a third option unfortunately. Sorry again, but what is delta? Is it the measurement recorded for each participants? As you can see, stats isn't my best subject.
 

Miner

TS Contributor
#7
Delta is the difference in pain threshold between the control leg and the treatment leg. Delta = treatment - control.

If you were to use the treatment threshold by itself, the variation in subjects pain thresholds will reduce the power of the test.
 

noetsi

Fortran must die
#9
I am not an expert in your area but there are many different approaches to do this depending on how you collected your data. One is do a test retest approach for A and B. Another is, depending on if someone got one or both treatments is two one way ANOVA (one for A and one for B) or a two way ANOVA.

Interval data is say 1 to infinity (a head index).
Ordinal is say something where you can order your data. So you have 3 levels and 1 is higher say on pain, than on 0 and 2 is higher than either on pain.
Categorical is something you can not order. Say religion with Catholic and Protestant being the measures. The simplest approach is dummy variables.

All use various methods (require various methods).
 
#10
Thanks everyone for your input. After spending all day on spss and trying different things, it looks like the independent t test between the means of treatment A and treatment B is the best option as I got a p value of 0.051. One-way anova between control, A and B gave me really high p values. Using delta between treatment A, B and their respective controls didn't give me anything significant either. I'm still not 100% sure that I fully understand the results given to me but there is progress.
 
#12
ANOVA
PostRx

Sum of Squares df Mean Square F Sig
Between Groups 3292.404 2 1646.202 1.509 .238
Within Groups 31626.515 29 1090.569
Total 34918.919 31
 
Last edited:
#14
Independent Samples Test
Levene's Test for Equality of Variances t-test for Equality of Means
F Sig. t df Sig. (2-tailed) Mean Difference Std. Error Difference 95% Confidence Interval of the Difference
Lower Upper
PostRx Equal variances assumed .408 .533 2.139 14 .051 28.575 13.360 -.080 57.230
Equal variances not assumed 2.139 13.382 .051 28.575 13.360 -.205 57.355
 
#17
1 1 8 Column on far left is participant number (17 onwards is the control legs).
2 1 144 Middle column is treatment (1 ice, 2 gel, 3 control)
3 1 140 Colum on right is measurements
4 1 166
5 1 114
10 1 120
11 1 143
16 1 81
6 2 68
7 2 86
8 2 109
9 2 135
12 2 83
13 2 87
14 2 122
15 2 74
16 3 71
17 3 97
18 3 72
19 3 169
20 3 76
21 3 142
22 3 176
23 3 99
24 3 76
25 3 74
26 3 105
27 3 178
28 3 94
29 3 104
30 3 117
31 3 81