Im testing the change in lesion volume between disease stages. I have measured lesion at T0 and T1, and calculated the change in volume with formula:
(VolumeT1 - VolumeT0) / VolumeT0 * 100
In my sample, there several subjects that has increase in volume (positive number) and several subjects has decrease in volume (negative number).
Can I just input all these mixed value into SPSS? and does these mixed values affect the distribution of my data set and affect the choice between parametric and nonparametric tests? Thank you so much
Yes, mixed values of a variable (both positive and negative) can affect the result of SPSS analysis. Here are some ways it can impact the analysis:
1. Mean and Standard Deviation: The presence of both positive and negative values can influence the mean and standard deviation of the variable. The mean can be pulled in either direction, and the standard deviation can be larger due to the spread of values.
2. Correlation Analysis: Mixed values can affect the correlation coefficient, as positive and negative values can have different relationships with other variables.
3. Regression Analysis: The direction and strength of the regression slope can be influenced by the mix of positive and negative values.
4. Hypothesis Testing: The results of t-tests, ANOVA, and other tests can be affected by the mix of positive and negative values, as they can impact the calculation of test statistics and p-values.
5. Data Visualization: Mixed values can impact the interpretation of plots and charts, such as scatterplots, histograms, and box plots.
To address these potential effects, consider the following:
- Data Transformation: Consider transforming the variable to a consistent scale (e.g., absolute values or standardization).
- Data Split: Split the data into separate groups based on the sign of the values (positive or negative) and analyze each group separately.
- Robust Statistics: Use robust statistical methods that can handle mixed values, such as the median and interquartile range (IQR).
Remember to carefully evaluate the research question, data distribution, and analysis goals to determine the appropriate approach.
The attached report is from https://copyleaks.com/ai-content-detector.
Onipe Adabenege Yahaya is at it again.
Please stop spamming questions you don't know the answers to.
Dear Bruce Weaver and Ronán Michael Conroy I would like to explain briefly my study design and variable, I also attach a small dataset showing original data.
In my study, I want to compare the change in lesion volume/density between 2 disease stages. I collected and measured data before (T0) and after (T1) treatment.
Variables:
- Change on lesion volume/density was calculated by: Δ=value T1 - value T0
- Percentage of change was calculated by: %Δ = (value T1-value T0)*100/value T0
In case lesion decreases in volume/density, the change in volume/density will appear as negative value.
I input these data into SPSS and run normality test: ΔV & %ΔD are normally distributed, %ΔV & ΔD are not normally distributed. Then I use independent t-test for ΔV & %ΔD, Mann Whitney test for %ΔV & ΔD. SPSS gives result as normal. However, as a beginner in research field, I am not sure if mixed negative and positive result can affect the result of SPSS.
Hello Huy Quang Nguyen. First, in the situation you described, there is no reason why you cannot have both positive and negative change scores. Second, and more importantly, relying on a statistical test of normality to determine whether you can use a t-test (or some other method) is a bad idea. Here is a short conference presentation I gave several years ago that explains why I say that:
Presentation Silly or Pointless Things People Do When Analyzing Data: 1. ...
In my view, if it is both sensible and defensible to use means and SDs for description, then you should encounter no major problems using a t-test. But if there are reasons to avoid using means & SDs for description, then you should consider some other type of analysis.
Finally, is "stage" the group variable for the independent groups t-tests you mentioned?
Dear Bruce Weaver,
First, I would like to thank you for your answer and your time
About your previous question, yes, the "change of volume/density" is my test variable. And "stage of disease" is my grouping variable
So my first question is clear: there is no problem to have both positive and negative value in test variable.
For my second problem, which is "which test to choose?". I don't understand this part "if it is both sensible and defensible to use means and SDs for description". So in my case, I want to describe and compare between 2 stage after treatment has completed, for example:
I want to find out after treatment, lesion volume in patient diagnosed with stage 1 reduced/increased by how many mm3 on average, and lesion from stage 2 patient increased/reduced by how many mm3 on average, and is there any significant difference between 2 stages.
If this is my target, then mean and SDs is sensible, isnt it? Then I can use independent-sample T-test even when my test variable is not normally distributed?
Huy Quang Nguyen, you need to bear in mind the following points.
- For an unpaired t-test, a sufficient normality assumption is that the two samples have been drawn from populations that are normally distributed.
- The necessary normality condition is that the sampling distribution of Xbar1 - Xbar2 is approximately normal. The shape of that sampling distribution depends jointly on the shapes of the underlying populations and the sample sizes.
- The normal distribution is infinite, ranging from negative infinity to positive infinity. Your samples are finite, and do not range from negative infinity to positive infinity. Therefore, your samples are not normally distributed.
I have no experience working with your two dependent variables, so I cannot tell you if it is sensible and defensible to use means and SDs as descriptive statistics for those variables. If you do not have the content knowledge and experience to make that judgement yourself, you'll have to consult some people who do have it, or consult the literature. Is it common to use means and SDs for those variables in the literature? Or do authors use medians & IQRs instead, for example?
Best of luck with your work.