What kind of SPSS analyses for data from transverse patterning experiment

I have to analyse data from a transverse vs. non-transverse patterning learning experiment, structured in two parts. All participants have taken part in both parts, so it is only a within-participants study (1. part non-transverse patterning and then immediately 2. part transverse patterning). Each part was structured in 3 phases in which the participant had to identify the "correct" symbol in a presented pair (1. phase AB, 2. phase BC, 3. phase AC and 1. phase AB, 2. phase BC, 3. phase CA resp). The participant (starting with phase 1) had to identify 14 correct pairs in a row to enter the next phase. My hypos so far:

Null Hypothesis (H0): Number of trial attempts for TP task will not significantly increase over NTP task.
Alternative Hypothesis (H1): Number of trial attempts for TP task will significantly increase over NTP task.

What kind of SPSS analyses would you recommend? Tbh I am not quite sure what exactly my independent variable is and how many there are. Is it only one IV, that is "learning or condition" with 2 levels (TP and NTP), or are TP and NTP two IVs? And how do I involve the 3 different phases in each part?
I guess my dependent variable is the number of attempts each participant needed.

I would be really thankful if someone can give me some advice!

Hello, Aya911, Kind of analysis is depends on number of participants and distribution of number of trial attempts. It may be t test for independent groups or (mosy likely) Mann-Whitney test. I would prefer to compare number of wrong attempts. Of course, you have only one IV with 2 levels, because a participant can't receive 2 kinds of conditions simultaneously. I think, better way to summarize attempts of all 3 phases for analysis. Also I think that it would be better to use crossover design for such experiment.
Hi ask.biostat,
Thanks for your reply! After eliminating a few outliers, there are 127 participants in the final sample. After checking distribution of total attempts for all 6 conditions, it seems that only one (TP3) is normally distributed. Do I have to perform a non-parametric test as the Wilcoxon instead of a t-test?
And yes, crossover design makes sense as it is kind of a repeated measure. Can an ANOVA be used to compare main effects of NTP and TP? If yes, would reckon a one-way repeated-measures ANOVA is suitable?
Hi, Aya911, sure, it is depended samples, sorry for mistake. How did you check the distributions? As I know, is it possible to use ANOVA for variables with "not quite normal" distribution. Make histograms. I suggests, the distributions have heavy left tail, so using log-transfomation should be suitable for normalizing of the distributions, after that you be able to use r.m. ANOVA.
Hi ask.biostat,
First I checked descriptives for all sets and noticed that skewness and kurtosis were quite high for all conditions except TP3. Then I made histograms (total number of attempts) where you could see the heavy left tail you mentioned. I also made histograms of numbers of wrong attempts, same here. But if I use log-transformation, then data should be suitable for t-test and ANOVA?
And does it make sense to present in my report descriptives of total attempts but distribution/histograms of only wrong attempts or is that confusing if I want to compare means of original data (total attempts) in the t-test later on? I hope I could articulate my thoughts correctly :)
Aya911, if you should use total attempts in your report, then use this variable in analysis. I think it would be better to provide r.m. ANOVA for log-transformated values, also I would make Wilcoxon test for raw data in order compare with ANOVA.
Hi ask.biostat,
I have lo transformed the data but it is still skewed. Shall I use Friedman instead? It is weird, I think according to our instructor we are supposed to perform an ANOVA, but I clearly demonstrate that my data is not normally distributed and thus parametric tests would not be suitable?
And what kind of analyses could I use to check whether there is an influence of gender and age on number of trial attempts? I have thought about a Spearman's rho for age, but what about gender?
Hello, Aya911! May you present the histograms? You may use ANOVA in case of not "heavy" tails. I agree with Spearman's rho for age. For gender, you may provide Wilcoxon test within transverse patterning and non-transverse patterning separately.