What can InPhonDB do for me?

Community-augmented meta-analyses are a new concept. So you may wonder, "what's in it for me?" Here are some answers - if you have other questions, let us know!

1. If you are thinking of gathering infant sound discrimination data
1.1. Has someone else already worked on the same contrast?
  1. Download the xls version
  2. Open in Excel
  3. Click on Data/Filter/Auto filter (libre/openoffice have similar menus)
  4. Scroll to the column entitled "contrast"
  5. Click on the little triangle/filter symbol on the right corner of that column
  6. Deselect "all", then select the contrast you are interested in

1.2. Which method gives me the lowest attrition for my chosen age group?
  1. Download the xls version
  2. Open in Excel
  3. Click on Data/Filter/Auto filter (libre/openoffice have similar menus)
  4. Scroll to the column entitled "mean_age_1"
  5. Click on the little triangle/filter symbol on the right corner of that column
  6. Click on "Standard filter" or "Special filter" or "Custom filter"
  7. In the next dialogue box, choose your condition to be that this column is younger than the maximum age (in days) your participants-to-be will have, and older than the lower age bound. (We use the conversion: month x 30.42 days/month). Press OK to apply the filter. You will now only see the rows within the age group that is relevant to you.
  8. You have several ways of getting average attrition rates across the different methods. Perhaps the easiest is to sort columns on the basis of "method", and then do averages by hand on n_exlcuded_1. (You can also do pivot tables, or read the csv data into your favorite analysis program …)
1.3. How large should my sample be for my chosen age group and method?
Here is an easy way to get a general idea of your target sample size based on sample sizes of previous work. 
  1. Download the csv version with effect sizes
  2. In a Mac, you can drag and drop it in Excel. Otherwise, you may have to click on the "File" menu in Excel, choose "Open…" navigate to and select the file you have downloaded, inphondb_ES.csv. Or you may have to click on the "Data" menu, then "Import"...
  3. In any case, when importing, make sure that "comma" is the selected separator, that UTF-8 is the encoding, and click OK to finish. You should now see the database -- the last column is called "Formulas.used", and everything should be aligned there. Check also that the column called "contrast_pseudoIPA" still contains unicode phonetic symbols. If you cannot see it, you may have to rely on contrast_sampa, and use the language-appropriate SAMPA to work back which International Phonetic Alphabet symbol it was.
  4. Click on Data/Filter/Auto filter (libre/openoffice have similar menus)
  5. Scroll to column P "mean_age_1"
  6. Click on the little triangle/filter symbol on the right corner of that column
  7. Click on "Standard filter" or "Special filter" or "Custom filter"
  8. In the next dialogue box, choose your condition to be that the column entitled "mean_age_1" is lower than the maximum age (in days) your participants-to-be will have, and greater than your lower age bound. (We use the conversion: month x 30.42 days/month). But don't press OK just yet!
  9. Add a second condition: that the "method" is your chosen method (see here for codes). Now you can click OK.
  10. Now you can get the mean or the median of columns "n_1" and "n_excluded_1". 
The following way is much better, though: You can do a prospective power analysis. 
  1. Follow steps up to 6 above, and adapt steps 7 and 8 once you have decided what are the "comparable" studies -- you SHOULD keep only studies with your chosen method, but other variables are left to your judgement. For instance, if I were doing a study on 6 mo using Central Fixation, then my filters could be: method="CF" AND 120 < mean_age_1 <=365 (so babies between about 4 months and 1 year). This selects 56 rows. You might then decide to exclude some more studies because they use unusual contrasts, or a very different design from yours. In my case, I left only studies on "native" contrasts (using a filter on the nativeness column).
  2. Once only relevant studies are left, copy the columns entitled x_diff, SD_pooled, and d into a new sheet
  3. Calculate the median values for these columns -- in my example, the median difference score is 0.91, the median pooled SD is 2.72, and the median effect size is 0.33.
  4. Use your favorite power calculation system/online tool to estimate sample size. I used the following formula in R (package: pwr): power.t.test(delta=0.91,sd=2.72,power=0.8,sig=.05,type=c("paired"),alternative="two.sided") to find out I needed to test 72 infants for my example case. If you don't like the number you get, you might consider concentrating on a contrast that gives larger effect sizes. See question 3.1 below.
2. If you have already gathered infant sound discrimination data
2.1. Is my effect size as large as that of comparable studies?
Let's say that you did a study on 6- to 8-month-olds using Central Fixation.
  1. Download the csv version with effect sizes
  2. In a Mac, you can drag and drop it in Excel. Otherwise, you may have to click on the "File" menu in Excel, choose "Open…" navigate to and select the file you have downloaded, inphondb_ES.csv. Or you may have to click on the "Data" menu, then "Import"...
  3. In any case, when importing, make sure that "comma" is the selected separator, that UTF-8 is the encoding, and click OK to finish. You should now see the database -- the last column is called "Formulas.used", and everything should be aligned there. Check also that the column called "contrast_pseudoIPA" still contains unicode phonetic symbols. If you cannot see it, you may have to rely on contrast_sampa, and use the language-appropriate SAMPA to work back which International Phonetic Alphabet symbol it was.
  4. Click on Data/Filter/Auto filter (libre/openoffice have similar menus)
  5. Scroll to the column entitled "mean_age_1"
  6. Click on the little triangle/filter symbol on the right corner of that column
  7. Click on "Standard filter" or "Special filter" or "Custom filter"
  8. In the next dialogue box, choose your condition to be that this column "mean_age_1" is lower than 243, and greater than 183. (We use the conversion: month x 30.42 days/month). But don't press OK just yet!
  9. Add a second condition: that the "method" is "CF", meaning central fixation. Now you can click OK.
  10. Now you can get the mean or the median of the column "d". 

3. If you work in a related research question
3.1. I'm interested in doing a minimal-pair word learning study, so I want to choose a contrast that 10- to 14-month-olds can easily discriminate. What contrast should I pick?
  1. Download the csv version with effect sizes
  2. In a Mac, you can drag and drop it in Excel. Otherwise, you may have to click on the "File" menu in Excel, choose "Open…" navigate to and select the file you have downloaded, inphondb_ES.csv. Or you may have to click on the "Data" menu, then "Import"...
  3. In any case, when importing, make sure that "comma" is the selected separator, that UTF-8 is the encoding, and click OK to finish. You should now see the database -- the last column is called "Formulas.used", and everything should be aligned there. Check also that the column called "contrast_pseudoIPA" still contains unicode phonetic symbols. If you cannot see it, you may have to rely on contrast_sampa, and use the language-appropriate SAMPA to work back which International Phonetic Alphabet symbol it was.
  4. Click on Data/Filter/Auto filter (libre/openoffice have similar menus)
  5. Scroll to column "mean_age_1"
  6. Click on the little triangle/filter symbol on the right corner of that column
  7. Click on "Standard filter" or "Special filter" or "Custom filter"
  8. In the next dialogue box, choose your condition to be that this column "mean.age.days" is lower than 426, and greater than 302. (We use the conversion: month x 30.42 days/month). But don't press OK just yet!
  9. We are going to remove CHT studies, because your word-learning study will use a method that's more similar to the others. So you add a second condition: that the "method" is not "CHT". Now you can click OK.
  10. Select the resulting table with 20+ rows, copy-paste into a new spreadsheet.
  11. Create a pivot table (there are plenty of tutorials out there about these) with average effect sizes per contrast. Looks like /e-i/ has effect sizes consistently above .6, so it's a good choice.
3.2. I'm interested in doing a distributional learning study, so I want to choose a contrast where performance in 10-month-olds is neither too poor nor too good. What contrast should I pick?
Follow the instructions in the previous question, except you'll choose a contrast close to the average effect size in the very last step. /o-u/ or /i-I/ might be good choices (effect sizes around .28).

3.3. I wonder if contrasts encoded in sonorous vowels attune earlier than contrasts encoded in non-sonorous vowels (thanks to Amanda Seidl for the question and the sonority coding!)
I'll do this one in R because it seems easier to me, but probably you can do something similar in Excel or your favorite stats program.
  1. Download the csv version with effect sizes
  2. In a Mac, you can drag and drop it in Excel. Otherwise, you may have to click on the "File" menu in Excel, choose "Open…" navigate to and select the file you have downloaded, inphondb_ES.csv. Or you may have to click on the "Data" menu, then "Import"...
  3. In any case, when importing, make sure that "comma" is the selected separator, that UTF-8 is the encoding, and click OK to finish. You should now see the database -- the last column is called "Formulas.used", and everything should be aligned there. Check also that the column called "contrast_pseudoIPA" still contains unicode phonetic symbols. If you cannot see it, you may have to rely on contrast_sampa, and use the language-appropriate SAMPA to work back which International Phonetic Alphabet symbol it was.
  4. You could code sonority yourself in an additional column. Alternatively, you could download a csv file with a sonority coding. This is a sample for add-on files: two columns coding the sonority of the first and second vowels in each contrast, and a column uniquely identifying each line. You can then copy and paste these columns in excel, or use this script to integrate extra columns into the database.
  5. Download this sample script (attention!! column names outdated) from the file cabinet; it contains instructions to make the plot below
  6. To make sure we're looking at controlled comparisons, we'll only include studies where multiple age groups were tested on the same contrast. Since there aren't very many non-native datapoints, we'll only look at native ones.
  7. To simplify exploration, we'll just look at contrasts where both vowels were coded as high in sonority (H H, in black, points in complex/filled symbols, regression line is solid and in black) and contrasts where both are low in sonority (L L, points in open symbols, regression line is dashed and in gray). The results are plotted below. 
  8. Although more sonorous vowels are discriminated slightly better than those with low sonority, the improvement in effect sizes with age is very similar in both types.
Effect sizes as a function of vowel sonority and age