Inferential Statistics Flashcards
What is the difference between positive and negative skewed distribution curve
Where frequency is not spread evenly (normal distribution curve where mean median and mode are the same) and data is clustered at one end. Positive is a slide starting at the top from the left and sliding down to the right into free space. Negative is a mirror image and the slide is straight into the y axis so is bad. For + mean is greater than Median and other way round for -
This information is in the extra bits pack which you don’t have so use the study directory!!
They’ve got practice qs as well
The difference between type one and type 2 errors
Type one is when the alternative hypothesis is accepted and null is rejected but behaviour was really due to chance (false positive) type 2 is when null is accepted / no sign diff when the iv was really having an effect (false negative)
What do these symbols mean: <,«,>,», wavy one these ~ and 3/4s of an infinity sign/a fish
Less than, much less than, greater than, much greater than, approximately, proportionately
What is the standard level of significance
P is less than one equal to 0.05 (less than 5% probability factors due to chance so accept alternative. If p is more than 0.05 there’s a more than 5% probability results are due to chance so accept null. If p less than or equal to 0.01 less than 1% p due to chanceso alternative is strongly supported
Draw out the table for when you would use each statistical test
Do it! Until you know it off by heart Or use the rhyme Carrots Best Mashed With Swede Under Roast Potatoes or cmu, bwr and sp-p. Remember to say that it is an EXPERIMENT with independent measures/repeated/matched ps
What is the criteria to use a parametric test
1.Have interval or ratio data, 2.have a normal distribution of data/scores (68% of data less than plus or minus 1 standard deviation from the mean),3. have similar variances between results from the different conditions (homogeneity of variance). -also no extreme score, continuous data E.g. related and unrelated t test or Pearson’s product movement NON PARAMETRIC have nominal or ordinal data
How to carry out the binomial sign test
- Indicate the direction of difference for changes in x e.g if more favourable put +, if negative out -. If unchanged leave black and DO NOT INCLUDE WHEN WORKING OUT NO. PS 2. Count the no. Times the least frequent sign (+or -) appears and call this the calculated value of x 3. Use critical values to see if there’s a sig dif. If calc lower or equal than critical, accept alternative
What is the rhyme to remember whether the calculated value should be higher or lower than the critical?
Chi is high, so is rho. All the rest are low (or equal to)
How to calculate chi square test
- Write down observed frequencies 2. Write down expected frequencies ( for each cell, cells row total x colum total divided by overall total) 3. 0-E 3. 0-E squared 4. 0-E squared divided by E 5. Repeat for each cell in the row 6. Add up new final column to give calculated value 7. Work out degrees of freedom (no. Rows -1 x no. Columns -1) to find critical value. If x higher or equal- accept alternative
How to carry out the Wilcoxon signed ranks test
- Ignore scores for S who scored the same in both conditions 2. Calculate the difference between conditions A and B for each score 3. Rank the differences ignoring + or -, give smallest rank of one and work up. If more than one have the same count up the scale how many have the same, find the middle one and give them all the score and move onto the next available number discounting the ones you counted up with. 4. Add the ranks together for both signs apparently?? And add up ranks for the least frequent sign = calculated value . For crit value n (discard those who got same)
How to carry out the Mann Whitney u test
- Place all the responses in rank order (rank both conditions together in one table) by having a table with the headings participant number, frequency (whatever they were measuring) and rank position. (Can use different colour pens if easier) Do ranking where lowest gets score of 1 and if more than one have the same e.g. 2, 2,2 count up 3 in the scale (2.3.4) and give them all the middle value (4) and the next rank gets 5. 2. Calculate 2 U values (one for each group) using the formula u1= sum of ranks for group 1 - no. People in group 1x (n+1) over 2 and same for u2. 3. Pick the lowest one which will be the calculated value. Critical depends on type of hypothesis and if beaten by a lot, go to next level of significance
How to calculate spearman’s rho
Rs= 1-6 times (sum of difference between pairs of ranks squared) over number of pairs times(no. Pairs squared -1) 1. Rank the scores for the 1st co variable 2. Rank scores for 2nd co variable (in separate columns) 3. Calc difference between rank A and rank B (A-B) 4. Square each values in the difference column, 5. Add up all d squared and plug into formula to get calc value
How to write a statement of significance
The calculated value of x is lower/higher than the critical value of y at 5% level of probability as p>/< 0.05. This means that there is/or no significant difference/relationship (for correlation) between R and S. Null/alternative hypothesis should be accepted HAVE TO SAY BOTH e.g. null rejected, alt accepted & vice Versa and for null:any diffs/results due to chance
How to write a conclusion from data
Need: point W main trend , result (work out mean/median/mode to back up), suggestion for why this may be or an application of the outcome . If correlation, state positive or negative or no but one doesn’t cause the other as correlation doesn’t show cause and effect , point out anomalies
